Every entry is dated and traced to a named source: 254 regulations, 173 deployments, 37 pieces of evidence of adoption and 8 pieces of evidence of impact.
More on Government & Public Sector across Europe lands here as we find it.
Add us as a preferred source on GoogleYour essential monthly update on Applied AI, with Government & Public Sector specially tracked.
Częstochowa City Hall — Invited by the Ministry of Digital Affairs, Częstochowa City Hall will join the AI HUB trial of an AI assistant for clerks built by NASK. Tasks: summarising and comparing documents, finding and analysing legislation with sources, anonymising files and helping prepare official letters. Staff review every output of this AI support for municipal staff; kick-off is 16 October.
Internetstiftelsen — Internetstiftelsen found 9% of Swedes sought AI voting advice before the 2026 election, and 22% of those said the answers strongly shaped the ballot cast. Composing political or civic texts with AI services has doubled since the start of the year, now two in ten. The study of chatbots for election information used 2,142 web interviews with people aged 18 to 84 between 17 and 27 September 2026.
Government of Italy (Legislative Decree 179/2026) — Legislative Decree 179/2026 appoints the Agency for Digital Italy (AgID) and the National Cybersecurity Agency (ACN) as EU AI Act watchdogs; ACN, Bank of Italy, CONSOB, IVASS and the privacy regulator split market oversight. From 23 October 2026, this Italy AI regulation aligns domestic law with Regulation (EU) 2024/1689 and covers AI in professional training. Decree 160/2026 (15 September 2026, AI Law 132/2025) earlier fixed the AI rulebook for policing and liability.
Seimas of Lithuania — The Seimas backed Social Democrat MP Birutė Vėsaitė's amendments to the Law on Public Information 51–5, with 11 abstaining, advancing Lithuania AI regulation. Sharing material made or reworked with AI tools that misrepresents real people's identity, words or deeds would be banned. These AI content rules go to the Culture Committee, then plenary on 10 December.
Spanish Data Protection Agency (AEPD) — The Spanish Data Protection Agency (AEPD) warned two city councils over AI image analysis in planned video surveillance, testing Spain AI regulation. One scheme includes facial detection or recognition, the other searching people by appearance. Under these AI surveillance rules, owning a technology does not justify using it.
Ministry of Digitalisation and Public Governance (Norway) — Norway's digitalisation ministry plans a temporary AI glasses ban at gathering spots such as beaches, parks, malls, museums and amusement parks, plus kindergartens, playgrounds, schools, youth clubs, children's events, gyms with changing rooms and health institutions. Minister Torgeir Micaelsen wants legislation before the Storting soon, warning wearers can photograph, film or record bystanders unnoticed. A specialist panel will later shape lasting national rules on camera eyewear, with exemptions planned for vulnerable groups.
Ministry of Digitalisation and Public Governance (Norway) — Norway's government is drafting a fresh national AI strategy, due in the first half of 2027, to replace the 2020 plan written under the Solberg government. Minister Torgeir Micaelsen says the new Norway AI policy will steer public bodies and private firms alike and shape cross-border cooperation over frontier models. Companies, universities and state agencies will be consulted before publication, as fast-moving AI made the old document outdated.
Ukrainian Air Force — Ukraine's Air Force reports robotic turrets downing jet-powered drones, Russia's Geran-5 included, with AI air defence doing the aiming. Machine vision steers each intercept and removes human error, spokesperson Yurii Ihnat explained. President Zelenskyy counted above 12 units guarding Kyiv, eight more due nearby, as drone defence automation spreads.
Polish Border Guard; Polish Armed Forces — Poland's Border Guard and armed forces are trialling Inteligentna Granica (Intelligent Border), AI border surveillance fusing camera, drone and sensor feeds across several dozen kilometres of Belarusian frontier under Operation Safe Podlasie. Software flags suspicious movement so operators can respond. Border Guard specialists co-built the AI monitoring tool with the Artificial Intelligence Implementation Center.
Swiss Federal Council — Switzerland's Federal Council adopted the report 'Increasing the Efficiency of Administrative Processes through Process Automation and Artificial Intelligence', answering Postulate 24.3582 Silberschmidt and setting rules for AI in public administration. Priority goes to translation support, case and application checks, data handling, business management workflows and back-office finance and procurement. Federal Administration units must keep the tools human-centred, trustworthy and sustainable, and a sharper inventory of algorithmic systems should speed up scaling.
Government of Russia (subcommission chaired by Deputy Prime Minister Dmitry Grigorenko) — The National Plan for the Implementation of Artificial Intelligence for 2027 to 2030 has cleared a government subcommission chaired by Deputy Prime Minister Dmitry Grigorenko. This Russia AI policy sets 170 federal project indicators in 16 industries for 32 agencies; a presidential commission reviews the AI plan next.
Cambridgeshire Constabulary — Cambridgeshire Constabulary is trialling Citizen Assist, an AI crime reporting tool from the National Police Chiefs' Council's Digital Public Contact programme. Residents describe non-emergency offences conversationally and check the finished account before sending. Assaults, threats, damage and theft come first; wider adoption of AI-assisted reports depends on results.
Ministerio de la Presidencia, Justicia y Relaciones con las Cortes; Tribunal Supremo — Spain's Ministry of the Presidency, Justice and Relations with the Courts deployed AI in the Supreme Court, letting First (Civil) Chamber legal teams extract parties, issues, cited provisions and arguments from appeals in seconds. The chamber logged 7,279 new cases in the first half of 2026; 100-plus staff trained in July. This AI case analysis runs on isolated, state-owned open-source systems.
Glavgosexpertiza of Russia — Glavgosexpertiza of Russia started trialling AI design review, a large-language-model assistant for its experts. Before formal examination, project files get scanned for likely defects and gaps, delivering AI checks for construction as a preliminary report.
South Yorkshire Police — South Yorkshire Police launched Bobbi, a website AI chatbot for policing that fields everyday non-emergency queries to ease pressure on 101, where over 1,000 calls arrive daily and peak waits pass 20 minutes. Answers come only from force-approved content, never general internet material, and the assistant cannot log crimes. Hampshire & Isle of Wight, Thames Valley and Humberside already use this virtual assistant.
Supreme Court of Latvia (Augstākā tiesa) — Supreme Court of Latvia senators and staff have searched case law with Judicata, an in-house AI legal search engine, since September. Queries never leave the court's own servers, because local models do the processing. Rather than answering questions or interpreting law, the semantic search pulls the closest passages from anonymised rulings, Constitutional Court decisions and translated European Court of Human Rights judgments.
Comune di Varese (Polizia Locale) — The Municipality of Varese switched on AI video surveillance in Piazza Repubblica, with three cameras flagging brawls or chases to the Local Police control room. Officers can replay sharp footage from just before each alarm. One unit is thermal, the smart CCTV runs after dark with loudspeakers, and no provincial capital tested it sooner.
Supreme Court of Finland; Åland Lagting — Åland AI regulation cleared review as Finland's Supreme Court held the Regional Act on Supervision of Artificial Intelligence within the islands' powers, with no bar to commencement. Passed by the Lagting on 17 June 2026 alongside amendments to sections 1 and 4 of the regional act on market control of certain products, this AI oversight law supplements EU Regulation 2024/1689.
Malta Tourism Authority — The Malta Tourism Authority launched a free AI tourism data platform combining MTA Traveller Survey answers and National Statistics Office arrival figures. Journalists, researchers, policymakers, tourism operators and the wider public can query both through an AI-powered interface. Released for World Tourism Day, the service spreads AI-powered tourism insights.
Austrian Federal Chancellery (Bundeskanzleramt) — Austria's Federal Chancellery will embed AI for civil servants inside ELAK, the Bund's digital case-file system. Phased rollout of KI im ELAK begins within weeks and ends by late 2026, letting officials query records, condense paperwork and get answers without switching tools. Sovereign LLM-as-a-Service capacity at the Federal Computing Centre (BRZ) hosts this public-sector AI, and GovGPT already reaches around 20,000 federal employees.
Kommunernas socialtjänst (KST) — Kommunernas socialtjänst (KST) is trialling Helen Assistant, a tool for AI interpreting in social services covering over 80 languages, in selected client meetings through 31 December 2026. Clients may decline, and staff use the AI translation for social workers only where suitable. KST will assess results before any permanent rollout from 1 January 2027.
National Assembly of Bulgaria (52nd National Assembly) — Bozhidar Bozhanov, Alexander Simidchiev, Elisaveta Belobradova and fellow MPs put a Law on the Use and Development of Artificial Intelligence before Bulgaria's 52nd National Assembly, a national AI law enacting the EU AI Act. A Minister of Innovation and Digital Transformation plus a national coordinator would lead; public bodies could adopt chatbots and automated paperwork. This AI legislation fines people 500–1,000 leva and firms 10,000–50,000 leva, and bars training on children's data.
Bundesagentur für Arbeit (Federal Employment Agency) — Klara, an AI voice assistant government hotline tool for Germany's Federal Employment Agency (BA), now answers roughly 18 million SGB III calls yearly. She greets callers, logs identity and basic needs, and in four pilot regions settles routine unemployment-benefit questions unaided. This telephone-support automation aims to shrink hold times, forwarding harder cases to staff with details already logged.
Government of Spain — Spain AI policy now follows the IA360 Plan, a 12-month programme of concrete steps Pedro Sánchez unveiled to capture economic gains while putting safety, trust and protection of children and other vulnerable groups first. Measures span compute and new models, cybersecurity, schooling, jobs, data centres and oversight, plus an AI voucher for SMEs and freelancers. The national AI plan wants over half of companies using the technology by 2030.
Swiss Federal Chancellery — Switzerland AI policy for federal offices follows a new Federal Chancellery two-year roadmap keeping six of eight planned measures after funding shrank to 2 million francs. Work centres on mapping process gains, an in-house generative AI system and legal groundwork for administrative use. A handbook and office marketplace were scrapped, with the current project database serving this government AI programme instead.
Supreme Court of the Russian Federation — Supreme Court chairman Igor Krasnov signed the Concept for introducing artificial intelligence technologies into court proceedings on 6 September, advancing Russia AI regulation. Judges keep the last word. These judicial AI rules target up to 95% of the court system using AI by 2030, with machine-drafted texts forming 30% to 50% of procedural documents.
Humppila municipality — Humppila has abolished its technical director and administrative director roles, leaving municipal manager Minna Ylikännö to cover part of the administrative brief with AI agents in local government. Those two deletions, tabled by the municipal board on 6 March, closed a 180,000-euro gap. Agendas and contract drafts are among the paperwork the software turns out. Ylikännö said on 18 September that town halls and employers must face municipal AI adoption trimming headcount.
Comune di Argenta — The Comune di Argenta switched on Civia on 18 September, an AI assistant for council services fielding queries about forms, cut-off dates and what the town has on. Type the query as you would speak it — a nursery place, Saturday's listings, a copy of a certificate — and Civia replies, 24 hours a day, across languages. Justomezzo, whose trade is public-sector messaging, built the municipal AI chatbot with machine-learning work from Momentum Digital.
Government of Portugal (Ministry for State Reform) — Portugal's government will pilot AI in urban licensing before year-end in a few technologically prepared municipalities, Coimbra included, Minister for State Reform Gonçalo Matias said. Building and urbanisation law in force since 1 October permits electronic licensing, opening the way for AI-assisted planning permits.
Immigratie- en Naturalisatiedienst (IND) — Two undisclosed chatbots, Robin and ChatIND, now handle AI asylum processing tasks for the Netherlands' Immigration and Naturalisation Service (IND), assisting roughly 230 workers with document summaries, translations and interview analysis since early 2026. Despite IND's stated transparency rules, neither tool appears in the country's official algorithm register. Critics among lawyers and researchers fear flawed outputs from AI-assisted asylum applications could distort decisions on refugee claims.
Ministry of Digital Transformation of the Republic of North Macedonia — A first draft national AI strategy from North Macedonia's Ministry of Digital Transformation is open for comment via the ENER portal, ahead of planned public hearings. Stefan Andonovski anticipates sign-off before 2026 closes, with the blueprint paving the way for a later artificial intelligence use law stressing responsible, values-driven deployment over restrictions on building the technology itself.
Regione Piemonte — Regione Piemonte has switched on Auxil.IA, a Csi Piemonte build putting AI for public administration into the hands of its workforce and town halls region-wide. Launched 17 September at the Grattacielo Piemonte alongside Paolo Zangrillo, Alberto Cirio and Csi Piemonte chief Pietro Pacini, it shortens files, weighs versions, writes up meetings, handles languages and parks everything in the Region's cloud. Councils may apply from 1 October, widening generative AI in local government.
Ayuntamiento de Ciudad Real — Spain's Ciudad Real City Council will buy a Municipal Artificial Intelligence Platform through a 447,700 euro, four-month contract. Staff are to apply AI for local government to resident enquiries, e-office casework and building permits, with the supplier configuring and linking these municipal AI tools.
Hellenic Court of Audit — Greece's Hellenic Court of Audit (Ελεγκτικό Συνέδριο) now runs an AI virtual assistant for judges, an AI-assisted judicial tool hunting through statutes and precedent to sort case files faster. Goal: shrink rulings from near 1,500 days to 600-650 days by end-2027. This came with a digital and energy revamp, moving filings and pre-contractual checks on public contracts fully online.
Government of Portugal (Ministry of the Economy) — Castro Almeida, Portugal's Economy Minister, will launch AI training for workers via an October trial near Leiria, eventually meant to reach roughly 100,000 employees nationwide. Some 100 staff from assorted local firms make up the opening cohort, a template for expansion. Separately, five or six Marinha Grande mould makers get coaching on applying artificial intelligence skills.
City of Munich and Munich Police — Munich police and city officials in Germany put over 50 AI cameras at around 35 Oktoberfest spots for AI crowd monitoring, catching dangerous crowding early. Fraunhofer Institute staff read density and movement from footage anonymously. A new hub lets police, fire, rescue and transport crews steer visitors through that crowd analytics.
Federal Government of Belgium (Minister Vanessa Matz) — Vanessa Matz, the federal modernisation minister, has handed a draft national AI strategy to cabinet colleagues and the federated entities. Belgium AI policy must now win over coalition partners and regions before this proposed national framework gets adopted.
Italian Republic — Decreto Legislativo 160, dated 9 settembre 2026, squares national law with Regulation (EU) 2024/1689, Europe's artificial intelligence rulebook, and marks the sharp end of Italy AI regulation. It prescribes how police may run AI systems and pins down who bears civil and criminal liability. Gazzetta Ufficiale, Serie Generale 214 carried the Italian AI law on 15 September 2026, biting from 30 September 2026, after the Council of Ministers cleared both adaptation decrees on 4 August 2026 under article 24 of Law 132/2025.
Märkischer Kreis — Since 11 September the Märkischer Kreis has picked up at any hour, passing callers to Emka, an AI telephone assistant for citizen enquiries. Emka steers people towards the right desk, paperwork or colleague, works 365 days a year and copes with over 30 tongues. Outside opening times, weekends and bank holidays included, the district's AI switchboard fields the call; by day the human counter answers first.
Gong — 50 of 78 leading public bodies in Croatia already use or test artificial intelligence while the Government does not, according to 'AI u javnom sektoru: Kako Hrvatska koristi umjetnu inteligenciju', the earliest systematic map of AI in the public sector anywhere in the country. Half of that group, 26 bodies, admitted to holding no internal rules governing its use. Gong compiled the survey between February and the close of August 2026 via freedom-of-information requests, logging 55 distinct AI solutions that run from Microsoft Copilot and ChatGPT to biometric identification, medical diagnostics and the anonymisation of court decisions. Sergej Zupanic, who wrote it, called the picture of government AI use available beforehand fragmented and limited.
State Customs Committee of Belarus — Belarus's State Customs Committee plans a customs AI application shifting from electronic to "intelligent" processing, applying machine vision to shipping and commercial paperwork attached to declarations. The system will read and cross-check these documents against the declaration, plus scan vehicle images to match each vehicle against its past crossings and similar vehicles, flagging discrepancies to officers. This border control automation targets document and vehicle verification, surfacing anomalies for human review.
Agency for Nature and Forests (Flanders) — Flanders' Agency for Nature and Forests says it will put cameras over protected land whose artificial intelligence reads the picture and picks out a rising plume of smoke. Validation runs through 2026 in North Limburg and other representative sites, with live trials this autumn and a staged roll-out from 2027 giving reserves AI wildfire detection. Minister Jo Brouns was blunt: every minute counts once a blaze starts, and early fire detection is the best chance of containing it.
Office for IT and eGovernment — Serbia's Office for IT and eGovernment will launch an AI assistant for passport appointments by late September 2026, automating identity card and passport bookings on the eGovernment portal. Director Mihailo Jovanovic told RTS users will ask, in natural language, for the nearest police station opening. Testing is underway, with plans to extend this conversational AI for government services to school and kindergarten enrolment.
French state (Ministry of Public Accounts) — The French state has pledged six million euros to Mistral AI across 2026 and 2027, buying AI for government administration across ministries. Les Échos broke the story on 2 September; the Ministry of Public Accounts confirmed a non-exclusive deal won by tender, with Mistral AI engineers posted into departments. Cybersecurity, fraud hunting and clearing bulk caseloads at the Ministry of Justice lead this public-sector AI push.
Supreme Court of the Russian Federation — Russian courts across several regions are piloting four large language models trained using judicial practice, Supreme Court Chairman Igor Krasnov says. This test of AI in courts draws on legal positions and practice-forming rulings. The Supreme Court also endorsed a court AI integration concept and created a Centre for Judicial Competences in AI.
Ministry of National Defence of Lithuania — Lithuania's Ministry of National Defence issued a roughly €5 million procurement for AI airspace surveillance, merging military sensors, radars and other agencies' feeds live. Defence Minister Robertas Kaunas expects stronger air-threat warning and response by state institutions from this AI drone detection network. Piloting begins early 2027, full sensor coverage follows within 2027, and NATO allies' platforms connect by 2028.
Chalmers University of Technology — Chalmers University of Technology researchers found in a preprint, not yet peer reviewed, that 9.4 per cent of motions tabled in Sweden's Riksdag in the 2025–26 parliamentary year had at least one paragraph judged to be written with generative AI. None of these uses was openly declared. The detectors were trained on Swedish and British parliamentary texts from before chatbots, and show the share climbing from near zero in 2021–22.
Swiss Parliamentary Services — Swiss federal parliamentarians are getting PIA, an in-house AI assistant, for the autumn session. Users include council members and their personal staff, Parliamentary Services employees and parliamentary group staff; for now it works best on committee documents rather than internet research. Swisscom runs the project, keeping operation and data storage in Switzerland, with up to CHF 150,000 budgeted for a roughly year-long pilot.
Human Resources Management Authority of Montenegro — Montenegro's Human Resources Management Authority and the Regional School of Public Administration opened Next Gen Gov: three-month, full-time AI training for civil servants that spans data analytics, programming and UI/UX design. Prime Minister Milojko Spajić wants the civil service AI upskilling to equip officials to lead digital transformation and apply AI responsibly.
Government of Sweden — Sweden AI regulation gained an inquiry under Dir. 2026:116: special investigator Lisa Englund Krafft will settle how the public access principle applies when agencies run artificial intelligence and external cloud services. She must map practice, decide if material in such tools counts as official records, and test whether rules for public-sector AI obstruct uptake, reporting by 15 November 2027.
Politie Gent — Politie Gent's new AI chatbot in policing answers residents 24 hours a day on the force's site, following months of trials. It greets visitors as the Ghent force's assistant, handling routine queries so officers, per the force, gain more room for genuine face-to-face contact — part of a wider shift toward AI in law enforcement.
Government of Hungary — Hungary AI regulation now falls under the Minister of Science and Technology rather than the Minister of Economy and Energy, after a new government decree strips market surveillance duties from the latter and hands them to the former. Government Decree No. 139/2026 (VIII.31.) makes the switch official, amending the earlier Gov. Decree 90/2026 (V.13.) through a freshly inserted Section 168/A. The reassignment had effectively already happened before this paperwork caught up, since the supervisory body's own website was already presenting itself as answering to the Ministry of Science and Technology. The decree simply confirms, on paper, which ministry now carries responsibility for policing artificial intelligence products in the market.
State Duma — Russia's first federal AI law took effect on 1 September 2026, after the State Duma passed it on 8 July. It governs AI development, deployment and use, and separates 'sovereign' models, made entirely by a Russian company and developed, trained and stored in domestic data centres, from 'national' ones that may use foreign open-source components. Both gain status only once the government confirms they fit what the law calls traditional Russian spiritual and moral values.
Český telekomunikační úřad — Czech AI regulation now has a dedicated enforcer: Český telekomunikační úřad, the country's telecoms watchdog, has opened a standalone unit to police artificial intelligence systems. Anna Hroudová heads the team, having moved over from the Confederation of Industry and Transport of the Czech Republic, joined by three colleagues covering law, technology and analytics. Once enabling legislation clears parliament, the unit inherits authority to enforce the European AI Act, the bloc's framework for EU AI Act implementation, giving the regulator real supervisory bite over AI in practice.
Comune di Pieve Ligure — Comune di Pieve Ligure has deployed 43 artificial intelligence systems for AI in local government, cross-checking municipal paperwork against regulations and flagging discrepancies to staff. Trialled across the general secretariat, local police and technical office, the tool won over staff, so the council funded a wider rollout using the ordinary budget. Small municipalities like this one, population roughly 2,500, rely on AI for municipal administration where single employees often run entire offices alone.
Armed Forces of Montenegro — Digital Adjutant, a new AI system for military command staff, was built in-house by Montenegro's Armed Forces for information analysis, legal drafting and document preparation. Minister Dragan Krapović said military IT specialists created it to ease administrative workload and free personnel for harder duties. Defence officials called it a domestic AI capability for national defence, keeping sensitive data under full digital sovereignty instead of foreign systems.
Albanian government (Smart City project) — Twenty schools in Tirana, Albania now run AI school security cameras under a Smart City initiative from Prime Minister Edi Rama. Rather than merely storing footage, the system spots and alerts staff to suspicious activity as it happens. The capital's 20 sites mark the programme's opening stage, ahead of any wider rollout of smart surveillance cameras in schools.
German Federal Government — Berlin now hosts a freshly launched AI safety institute: AISI Germany, folding BSI's cybersecurity expertise together with the Federal Network Agency's former remit over operational safety of AI systems, all under one national AI safety body. A June 2026 resolution from the National Security Council brought it into being, with joint command shared between the Federal Ministry of Digital and the Federal Ministry of the Interior. Holding no sanctioning power, it instead weighs each AI model's dangers against its upsides, steering government, administration and industry as an analytical hub.
Ministry of Defence of Ukraine — Ukraine's Ministry of Defence has opened its military AI assistant to everyone on the DELTA combat system, trimming web-widget help tickets by roughly 27% and shortening replies by a factor of 30. Hosted inside DELTA's protected environment, this combat-system AI helpdesk tackles routine enquiries on-screen and fields around one in four widget messages.
MeteoSwiss — Switzerland's MeteoSwiss has activated an AI thunderstorm forecasting system after a four-year build. It rescans nationwide radar and lightning readings every five minutes, flagging storm-hazard odds within the following hour. This deep learning storm prediction engine trims false alarms and lengthens warning lead times versus MeteoSwiss's prior setup.
Government of the Slovak Republic — Slovakia has approved a draft act placing state administration of artificial intelligence under the Ministry of Investments, Regional Development and Informatisation (MIRRI), naming it the single contact point and central market surveillance authority for AI systems. This Slovakia AI regulation splits sector-specific duties among the National Security Authority, the Data Protection Office and the Justice Ministry, and requires MIRRI to build a regulatory sandbox for developing and testing AI under the new AI oversight law. The law takes effect 1 January 2027.
Odašiljači i veze — Odašiljači i veze has widened OIV Fire Detect AI, Croatia's AI wildfire detection grid, by 50 watch cameras across 25 fresh sites, favouring protected land such as national parks and nature parks; the estate now stands at 143 locations and 282 cameras. Advanced algorithms pick out smoke by daylight and flame after dark, engineered for unbroken cover 24 hours a day, seven days a week, year-round, with every county fire monitoring centre able to pull pictures and readings. Procurement fell to the Ministry of Environmental Protection and Green Transition; the forest-fire detection network was paid for through the Competitiveness and Cohesion Programme 2021-2027.
AADE (Independent Authority for Public Revenue) — Greece's Independent Authority for Public Revenue (AADE) has activated dedicated units for AI tax audits, running risk analysis and automated cross-checks. They are meant to improve how the authority spots tax risk and audits.
KPRM, Senate and Prokuratoria Generalna — Testing of an AI assistant for public administration is underway at three Polish bodies: the Chancellery of the Prime Minister, the Senate Chancellery and the Prokuratoria Generalna (General Prosecutor's Office). Deputy Digital Affairs Minister Dariusz Standerski described three planned uses: answering general questions on central government, a legal function that cites regulations plus relevant court decisions, and automatically anonymising documents. Built on the PLLuM language model, this AI tool for government officials should see two or three functions rolled out into central office systems by end-2026.
UK Home Office — The UK Home Office now uses AI to triage and route police 101 non-emergency calls, aiming to cut waiting times. The step belongs to wider work to reduce pressure on non-emergency phone lines.
UK West Yorkshire Police — West Yorkshire Police deployed a live facial recognition van in Bradford city centre, arresting four people wanted for crimes, finding a missing person and flagging 21 people under Sexual Harm Prevention Orders for compliance checks. It was the force's seventeenth such Bradford visit since February 2026, with 29 arrests in all.
Public Procurement Agency (Albania) — AI procurement remains stalled a year after Diella was made Albania's Minister of State for Artificial Intelligence in September 2025 to strip corruption from public tendering. Officials have reached only the paperwork stage, finishing the terms of reference a tender for the system itself requires before bidding can open. Agency director Reida Kashta confirmed work on the state artificial intelligence minister's promised procurement platform continues, while the national strategy and a parallel AI law both sit in consultation.
e-Albania (National Agency for the Information Society) — Albania has named four systems now driving AI in public administration through its e-Albania platform: the Legal Overlap System, which flags clashing or duplicated wording across legal acts; a module speeding translation and alignment of Albanian law with EU law inside the European Integration and Membership Platform; eIAD, which mines tax data for anomalies and risk; and Diella, which points citizens toward services and paperwork. Officials stress human staff still make the final call, though no launch dates were given for any of these four AI systems.
Government of Serbia — Serbia's Ministry of Science, Technological Development and Innovation aims to pass the country's first Serbia AI law by December 2026. Modelled on the EU AI Act, the draft sorts systems by risk into prohibited, high-risk, medium-risk and minimal categories, backed by a new AI Agency and a registry for the riskiest tools. The resulting Serbian AI regulation would give the country its own dedicated framework for governing AI use.
Ministerio de la Presidencia, Justicia y Relaciones con las Cortes — Spain's Ministry of the Presidency, Justice and Relations with the Courts reports saving over one million hours through AI in courts and robotic process automation, which completed 350,000 administrative tasks and handled procedures worth more than €700 million. Court automation tools span automatic transcription, data anonymisation, summarisation, document classification, case-file information extraction and AI-powered translation for citizen services.
UK British Transport Police — British Transport Police has taken its live facial recognition trial into London Underground stations, deploying NEC's NeoFace M40 at Victoria after starting at London Bridge in February. Concourse cameras check passengers against a police watchlist for high-harm offenders, especially people sought for sexual and other serious crimes. BTP says non-matches are deleted immediately; the pilot, first set at six months, now runs to November 2026.
Urząd Ochrony Danych Osobowych (Personal Data Protection Office) — Poland's Personal Data Protection Office (UODO) issued four checklists guiding Polish organisations through AI compliance checks under GDPR before building or deploying AI systems. Lists suit smaller firms, government bodies and others, plus a longer version covering EU AI Act duties like risk tiering and rights-impact reviews, supporting GDPR compliance for AI systems.
Protecció Civil — Protecció Civil wrapped up the wildfire-prevention drive with an AI-generated image showing the Madriu valley ablaze. Nothing in the scene is real; the body released the fake picture so Andorra grasps how close an actual blaze could be.
Consumer Protection and Technical Regulatory Authority (TTJA) — Under Estonia AI regulation, the Consumer Protection and Technical Regulatory Authority (TTJA) will serve as competent authority overseeing AI Act compliance. Its explainer covers in-scope goods, sectors and eight prohibited AI practices. TTJA may demand providers' training datasets, logs and source code under this national AI rulebook.
Government of Slovenia / AKOS — AKOS now enforces AI transparency rules in Slovenia under Article 50 of the EU Artificial Intelligence Act. Providers and deployers of AI systems that interact with people, or generate synthetic content, must flag the AI's involvement or the output's artificial origin. AKOS, Slovenia's Communications Networks and Services Agency, is also the country's AI Act disclosure obligations contact point.
Czech Telecommunication Office — From 2 August 2026 the Czech Telecommunication Office, among the market surveillance bodies the Czech Republic has designated, spelled out how the AI transparency rules of the AI Act bite. Anyone supplying a chatbot must tell people they are dealing with a machine, and supplier output carries a machine-readable mark; operators label only deepfakes and AI-made material touching the public interest. Biometric and emotion-recognition kit obliges its operator to notify every person subjected to it, and enforcement of Article 50 can fall to the European AI Office instead.
Government of Malta — Malta's new Artificial Intelligence Governance Council, led by Chief AI Officer Jonathan Gerada, now steers government-wide AI adoption in the public sector. The council unites public administration representatives to turn oversight decisions into concrete action, prioritising public value over cost reduction. This reflects Malta's approach to AI governance in government, favouring coordinated implementation over savings-driven deployment.
Ministry of Justice of the Republic of Serbia — Serbia's Ministry of Justice has drafted a Law on Personal Data Protection setting first-time AI data protection law rules for handling personal information within artificial intelligence systems, part of the country's broader AI governance regulation push. High-risk systems require an impact assessment plus the Commissioner's opinion before use; minimal-risk tools face no such demand. Synthetic audio and video need clear labelling, with public comment open until 10 September.
German Federal Government — Germany AI regulation now runs through KI-MIG, the AI Market Surveillance and Innovation Promotion Act, live since 29 July 2026 after Bundestag passage on 11 June and Bundesrat clearance on 10 July. BNetzA leads as watchdog, EU AI Office contact, and complaints handler under the EU AI Act. A built-in sandbox backs AI Act enforcement Germany, letting firms trial systems under supervision.
Government Office of Estonia — The Government Office has opened a five million euro tender for AI skills training, seeking at least 100,000 working-age residents through one Eesti.ai workshop or learning bite during 2026 and 2027. Each participant takes two four-hour sessions, 75 percent plus face to face and offered in every county, open to workers from 18; some 65 percent of places go to existing AI users, 30 percent to beginners and five percent to the advanced. A three-month pilot ran April to June before this AI literacy programme, which sits under "Kõige AI-targem rahvas" and drew 5.2 of the 11 million euros Eesti.ai won from the June supplementary budget, after the Estonian government's January start.
Polish Parliament and President of Poland — Poland's Sejm passed the Act on Artificial Intelligence Systems on 11 June 2026, the country's implementation of Poland AI regulation under the EU AI Act; President Andrzej Duda signed it into Polish law on 27 July 2026. It sets up the Commission for Development and Safety of Artificial Intelligence (KRiBSI) as market surveillance authority, creates regulatory sandboxes, and mandates AI-generated content labelling. The AI Systems Act Poland rules apply in stages from July 2026.
Minsk City Executive Committee — Minsk will add AI contact centre agents to the 115.bel housing hotline before the year is out, said Maksim Bardovich, who directs the Minsk City Executive Committee's Information Technology Centre. The agent-bots should shorten how long a request takes to handle and an answer to draft, sparing residents a queue. The hotline logged over 850,000 reports in the first half of 2026, routing them to city services.
Ministry of Labour, Employment, Veteran and Social Affairs — Serbia's Ministry of Labour, Employment, Veteran and Social Affairs put its first AI assistant for social protection queries, Radoslav, live on the eGovernment portal on 23 July 2026, running 24/7 without collecting personal data. The tool answers, conversationally, questions on qualifying for child support, family allowances and other welfare payments, plus application steps. Running on domestic infrastructure, including the Kragujevac supercomputer, the AI-powered social welfare chatbot clarifies rules rather than ruling on claims, leaving final decisions to human caseworkers.
Stadt Rüsselsheim am Main — Rüsselsheim am Main will introduce AI minute-taking from notes or audio, a semantic search and dialogue tool spanning archives, meeting papers and project documents, and an email assistant that sorts and ranks messages, part-drafts replies and condenses long threads. Hesse's state secretary for digital affairs handed the mayor 309,000 euros for the project under the Starke Heimat Hessen programme. The three tools will run on a German-based sovereign public-administration platform and later be reusable by other municipalities.
Ministry of Digital Transformation of Ukraine — Ukraine's Ministry of Digital Transformation, with Ukrainian developer Kitsoft, has added a voice-based government AI assistant feature to Diia.AI. Citizens can now speak queries and hear spoken replies instead of typing, and can switch between voice and text mid-conversation without restarting. The rollout extends this voice-enabled government AI service to make public administration accessible without typing or literacy barriers.
Oireachtas (Irish Parliament) — Ireland's Oireachtas has handed the country a dedicated AI regulator: the Regulation of Artificial Intelligence Act 2026 (Number 31 of 2026) creates the AI Office of Ireland, Oifig IS na hÉireann, tasked with enforcing the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) locally. This Irish AI regulation assigns market-watchdog duties to named bodies, opens sandboxes for testing high-risk systems under real conditions, and rolls out a public register, one-stop contact point, complaint-handling steps and penalties, together forming the state's new AI oversight framework.
Spanish Data Protection Agency (AEPD) — The Spanish Data Protection Agency (AEPD) has issued Spain AI regulation guidance: a technical note on data accuracy, suitability and quality where Artificial Intelligence processes personal data. This AI regulatory guidance examines how GDPR accuracy and minimisation principles bear on AI systems, and recommends steps to secure data quality for AI training and operations.
Austrian Federal Chancellery — Austria's Federal Chancellery has begun rolling out GovGPT, a sovereign AI assistant for civil servants, starting in the Chancellery and moving department by department. Staff draft texts, condense long documents and build searchable knowledge libraries, with queries and uploads processed on Federal Computing Centre infrastructure instead of a public chatbot. The Ministry of Finance followed from 21 July and the consumer protection, care, health, social affairs and labour ministry on 28 July, targeting 180,000-plus federal employees.
Hellenic Parliament — Under Law 5321/2026, Greece rolls out Greece AI regulation via EU AI Act implementation for Regulation (EU) 2024/1689. Oversight splits between the Hellenic Data Protection Authority for market surveillance and the Hellenic Telecommunications and Post Commission as notifying authority, alongside a sandbox, an AI Observatory and a public-sector systems registry. Stripping transparency labels from deepfake content now carries criminal liability.
Albanian Ministry of Interior (State Police) — Albania's Ministry of Interior has switched on AI policing in Albania spanning 873 sites: 12 Regional Police Directorates, 24 major cities, 28 border crossings and 20 Tirana schools all feed one Smart City command and control centre. Software there flags behaviour it judges dangerous or unlawful and raises an alert for officers to act on, and Interior Minister Besfort Lamallari insists the system amounts to more than cameras.
INPS (National Social Security Institute) — Italy's national social security institute, INPS, lists 59 AI systems classified under AgID guidelines in its 2026 annual report, bringing AI in social security to a multi-agent conversational chatbot and certified-email routing. The routing system automatically directed 70.4% of 5.4 million messages analysed from April 2021 to April 2026 with 87% overall accuracy, and generative AI helps staff check reimbursement documents in the Estate Inpsieme welfare programme.
Consell General — The Consell General passed Penal Code revisions without dissent, tightening Andorra AI regulation: buying child sexual abuse imagery online, AI-generated included, draws prostitution-level penalties. Those AI content rules join a consent age lifted from 14 to 16. Justice and Interior Minister Ester Molné tabled the qualified laws.
Royal Gibraltar Police — The Government of Gibraltar has brought Live Facial Recognition into round-the-clock service along the reopened frontier and further priority sites, with the Royal Gibraltar Police as data controller. A heavy volume of stills ran through AI face matching in week one, flagging nobody on a police watchlist. Recordings are kept to prevent and detect crime and protect public safety and national security. Chief Minister Fabian Picardo rates the technology estate, facial recognition included, as holding up well.
Government of Iceland (Ministry of Finance and Economic Affairs) — Microsoft Copilot is spreading through Icelandic state institutions under a three-year Microsoft agreement with the Ministry of Finance and Economic Affairs, driving a coordinated push for AI in government services. Officials cite better public services, stronger staff, greater efficiency and responsible artificial intelligence in the public sector as goals. Rollout speed depends on each institution, with large-scale training starting autumn 2026.
Autoriteit Persoonsgegevens (AP) — Two new instruments from the Dutch Data Protection Authority, AP, tackle AI GDPR compliance: technical guidance for developers and a self-assessment checklist letting organisations judge their own readiness before rolling out generative AI. Lawful bases, required protections and record-keeping duties under the GDPR run through both. AP separately revived its earlier consultation opinion on Microsoft 365 Copilot, tying personal data protection in generative AI back to that regulatory approach.
Regierungsrat, Canton of Zurich — The Canton of Zurich's government says it will make two piloted tools permanent across its administration: TranscriboZH, for AI transcription of audio and video, and KlartextZH, a plain-language text simplifier. Decided on 9 July 2026 (Regierungsratsbeschluss Nr. 749/2026), the step belongs to SynerKI, a cross-departmental project that will also build a central AI service platform for handling protected data. The Regierungsrat approved 2.76 million francs one-off and 3.87 million francs a year in running costs.
Parliamentary Assembly of Bosnia and Herzegovina — The Parliamentary Assembly of Bosnia and Herzegovina has cleared ratification of the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, the earliest binding treaty anywhere on using artificial intelligence. Bosnia and Herzegovina AI regulation gains no home-grown statute: consent, carried in the Official Gazette - International Treaties number 5/26 of 9 July 2026, ties the state to an outside legal framework that institutions and later rules must deliver. Signature came on 9 December 2025 in Strasbourg, settling the terms of national AI oversight.
Luxembourg Ministry of Digitalisation — Luxembourg's Ministry of Digitalisation put an AI chatbot government services tool live on Guichet.lu, the country's citizen portal, on 6 July 2026, under the AI4Lux campaign led by Minister Stéphanie Obertin. Residents can now type queries in whichever language they speak and get routed automatically among 2,000 bureaucratic procedures, rather than hunting for the right form by name. The assistant, an early example of conversational AI for public administration, runs entirely on government-owned servers.
Câmara Municipal de Portimão — Portugal's Portimão city council says it will adopt a digital planning platform, and its online service will use AI to check documents the moment residents submit them, as all building licensing moves online. Municipal staff training has begun, and the platform is arriving in phases, expected to go live within months.
Dutch government — Netherlands AI policy now includes an International AI Strategy, with the Dutch government setting out global cooperation on trustworthy AI. Three priorities anchor it: diplomacy on European and worldwide regulation, alliances safeguarding shared values, and stronger European AI infrastructure cutting reliance on outside powers, part of its Dutch AI strategy push.
Albanian Government (Prime Minister Edi Rama) — 140 senior civil servants across Albania's ministries have joined an AI training programme civil servants will rely on for sharper judgement and quicker decisions, Prime Minister Edi Rama said at Public Administration Week's third edition. This initial batch will grow into wider AI civil service training reaching other state bodies.
Skatteverket and Försäkringskassan — Sweden's tax agency Skatteverket and social insurance agency Försäkringskassan have begun trial operation of AI-verkstaden, a shared AI workshop that helps agencies, municipalities and regions build and use AI. Ten pilot organisations joined first, including Region Halland and Mörbylånga municipality, testing redaction, transcription and digital-assistant AI services. The pilots were picked to span different sizes and levels of technical maturity, the agencies said.
State Election Commission — Croatia's State Election Commission has opened GlasAI, an AI chatbot for elections fielding public questions on ballots, referendums and the commission's remit, built with the Ministry of Justice, Public Administration and Digital Transformation. Reachable at izbori.hr under a stylised ballot-box-and-headset icon, the electoral assistant has replied 24 hours a day, 7 days a week since 1 July 2026.
MASGZ Brandenburg — Brandenburg activated an AI in immigration services tool within 'Aufenthalt Digital' from 26 June 2026, expanding a six-month Dahme-Spreewald pilot statewide. The system flags missing or contradictory paperwork before submission, spanning permits for employment, training, family reunification and settlement. Some 350 German offices already run the portal, now enhanced with automated document verification for visa applications.
Ministry of Industry and Trade of the Czech Republic — A single point of contact for Czech AI regulation would sit with the Czech Telecommunications Office, tasked with leading market checks under a first-of-its-kind national statute now written by the Ministry of Industry and Trade to fold the EU AI Act into the law of the Czech Republic. Two other bodies, the Czech National Bank and the Office for Personal Data Protection, would supervise their own fields, and a sandbox for trialling artificial intelligence tools would open under the Czech Agency for Standardisation. Other ministries gave feedback in autumn 2025, and the bill went to cabinet on 26 June 2026; come September 2026 no minister had reviewed it yet, leaving this proposed AI Act Czechia measure stalled before reaching parliament.
Tallinn Administrative Court — Ten people challenging a felling permit were fined by Tallinn Administrative Court after filing a submission whose citations named Estonian scholars who do not exist and studies nobody could trace — the first penalty of its kind under emerging Estonia AI regulation. One paid 150 euros and nine paid 50 each, well inside the 3200 euro ceiling; Tallinn Circuit Court upheld the ruling and it became enforceable once the Supreme Court refused the appeal. Judges accept that parties may lawfully use such tools, yet hold them answerable for whatever they assert. Kaupo Kruusvee, who chairs the administrative court, would rather litigants wrote in plain words of their own, and while earlier warnings and reprimands over careless AI use in court are on record, no previous fine is known.
Crisis Management Center of North Macedonia — North Macedonia's Crisis Management Center (CMC) has agreed with NATO to implement NICS-2, a multi-year AI crisis response project delivering an incident command system, AI-driven, for cross-border emergencies. Under NATO's Science for Peace and Security Programme, this AI-assisted emergency response effort targets stronger national crisis management capabilities via AI-enhanced coordination and real-time decision-making.
Metropolitan Police — Live facial recognition policing expands to London's West End and Soho by Christmas 2026, the Metropolitan Police says, fixing cameras to street furniture ahead of six further rollouts in 2027. A prior Croydon trial scanned 470,000 faces over six months, yielding 173 arrests and one false match, evidence now cited to justify wider facial recognition surveillance.
Bundesagentur für Arbeit (Federal Employment Agency) — Germany's Federal Employment Agency (BA) is rolling out AI customer service for public agencies nationwide, extending its Antwortvorschläge reply-suggestion tool to every SGB III Service Centre, with completion due by 30 September 2026. The system drafts responses to mailbox queries, but staff must check and approve each one before sending. This nationwide push follows a longer internal test phase, with automated drafting still requiring human sign-off throughout.
Office for IT and eGovernment of Serbia — Serbia's Office for Information Technology and Electronic Administration teamed up with Microsoft to deliver AI-assisted public administration training to 1,500 civil servants using Microsoft Copilot. Lessons ranged from quicker document searches, summaries and drafting to handling routine bureaucratic duties, with government use of Microsoft Copilot extending to advanced modules in Word, Excel, PowerPoint and Outlook for select staff. The curriculum closed with building AI agents for workflow automation and instruction on responsible, data-protective AI practices.
Direction générale de l'administration et de la fonction publique (DGAFP) — Talks on AI in the civil service began in France as the DGAFP met unions and employers across all three public-sector branches to negotiate a framework agreement on rolling out and using artificial intelligence in government work. The draft, tied to the 'Notre IA' plan, groups its provisions into six strands: guiding principles, ethics and transparency duties, oversight structures, dialogue with staff representatives, workforce training, and handling organisational change from AI deployment in public administration. Negotiators are due back at the table on 10 July 2026.
Ministry of Science, Technological Development and Innovation of Serbia — Serbia AI regulation is taking shape: a working group under the Ministry of Science, Technological Development and Innovation aims to finish the nation's inaugural artificial intelligence statute by December 2026. Systems face sorting into four bands—banned outright, high-risk, needing disclosure, or minimal concern—while officials plan a new AI Agency alongside a registry tracking high-risk tools. The European Union's AI Act supplies the template, tying into Serbia's own AI Development Strategy stretching to 2030.
Eesti.ai council, Government Office of Estonia — Estonia's Eesti.ai council, set up at Prime Minister Kristen Michal's initiative, agreed at its second meeting to create a digital identity for AI agents, which it calls an AI personal identification code. Michal backed letting agents act for a person, company or institution within defined, verifiable and auditable rights, such as only viewing data, preparing a document or paying up to a set ceiling, never all of a user's rights. Moving quickly, he said, would make Estonia the first country to grant AI agents an official digital identity.
Dorset Council — Faculty, an AI firm working with the UK government, built the agents now being piloted by Dorset Council to assist planning officers on minor applications, reviewing documents and issuing recommendations. This AI planning agents trial aims to halve average processing times, cutting the wait from eight weeks to four. Two other councils are running the same test, with AI in planning application processing meant to ease backlogs.
French government (Notre IA plan) — The French government will bring generative AI for public advisers to its 8,000 France services staff by autumn 2026 with a tool built on Aria, a chatbot from the Directorate of Legal and Administrative Information (Dila). It is part of the Notre IA plan unveiled at Bercy, which also foresees a general conversational tool for some one million state employees, phase one costing an extra €750,000 in Mistral licences and computing power.
Federal Ministry for Digital and State Modernisation (BMDS) — Germany's Federal Ministry for Digital and State Modernisation has open-sourced SPARK Workflow, an AI planning approvals system any government level can now adopt. Agents flag gaps or contradictions in submissions and draft rulings pending human approval. Trials run at the Federal Highway Authority, Hamburg's hydrogen core grid permitting, and North Rhine-Westphalia's emissions clearances, pointing toward wider AI-assisted permitting.
Varsinais-Suomi and Satakunta wellbeing services counties — The Varsinais-Suomi and Satakunta wellbeing services counties launched joint work in June 2026 to take AI record-keeping in social care from pilots to full social welfare services. Over 600 Varsinais-Suomi professionals created 30,000-plus AI-assisted records in the pilot, about 300 in Satakunta produced over 10,000 in 2026, and a tenfold rise in use is targeted by end-2027.
Dutch central government (Ministry of the Interior and Kingdom Relations) — The Dutch government is extending vlam-chat, a home-built AI chatbot for civil servants, to several units of the Ministry of the Interior and Kingdom Relations, with pilots elsewhere in central government. The ministries of Health, Finance, General Affairs, Education, and Infrastructure and Water Management, plus the High Councils of State, already have it. 82% of surveyed users use it at least weekly and 77% save noticeable time; production rollout follows in the second half of 2026.
North Rhine-Westphalia tax administration (Finanzverwaltung NRW) — North Rhine-Westphalia's tax administration is live-testing FinAssist.NRW, an in-house AI assistant hosted in its own data centre. Tax offices in Gütersloh, Oberhausen-Nord and Solingen have trialled the chatbot live since June 2026, before the state finance ministry's own trial; it condenses and analyses internal papers, writes drafts, organises email chains and pulls out task lists.
Nagyoroszi Council — Nagyoroszi's council in Hungary has opened a freely reachable AI customer service, supplied by Munipolis, still learning and fed solely by the village's own web pages. Preset bubbles cover the commonest queries, typed messages work too, and adjacent topics appear unprompted. The municipal chatbot copes with several languages, spots the edge of what it knows and hands anything past that to whichever council officer owns the case; officials report lighter paperwork already.
BMDS — Complex GDPR data requests now clear in hours instead of weeks, a 90% speed gain logged by Germany's Federal Ministry for Digitalisation and State Modernisation across Agentic AI Hub trials in 19 municipalities. Care benefit applications moved 45% faster and staff spent 78% less time chasing up meeting minutes, as the scheme piloted AI agents in local government spanning 20 administrative workflows. 70% of those pilot solutions, an early marker of public sector AI adoption, were already live in daily operation before the cut-off.
National Cyber and Information Security Agency (NÚKIB) and Office of the Government — NÚKIB (National Cyber and Information Security Agency), with an inter-ministerial group under the Office of the Government, gives officials a Desatero — Ten Commandments — for using AI responsibly. Security, data protection, transparency and trust anchor its Guide for Ethical and Responsible Use of AI in Public Administration. Goal: safe, purposeful AI in government across Czechia's public sector.
Ministry of Digital Transformation of Ukraine — A new panel convened by Ukraine's Ministry of Digital Transformation held its inaugural session on 15 June 2026, tasked with producing AI regulation Ukraine can align to the EU AI Act by year's end 2026, ending the current reliance on voluntary self-regulation. Officials, business figures and civil society representatives jointly guide the resulting Ukrainian AI legislation.
Polish Civil Service (Służba Cywilna) — On 10 June 2026, Poland's Civil Service launched a third free e-learning course on AI for government administration, hosted on the Civil Service e-learning system. Content spans core generative AI concepts, secure usage habits, crafting effective prompts and applying AI responsibly, forming part of Poland's wider push on AI in public administration for its staff. Two prior modules on the same platform came before it.
Ministry of Justice, Public Administration and Digital Transformation — Croatia's Ministry of Justice, Public Administration and Digital Transformation has switched on an AI virtual assistant for public services at the heart of the state portal gov.hr, delivered inside the National Recovery and Resilience Plan scheme building one contact hub for every e-public service. Residents phrase a query the way they would put it to a clerk across a counter — renewing a passport, enrolling a child, registering a move — and the government service chatbot hands back an ordered reply, pointing at the matching e-service. Availability is 24 hours a day, 7 days a week, with no personal data demanded or retained.
Arcom (Autorité de régulation de la communication audiovisuelle et numérique) — France's media regulator Arcom now binds all 2027 presidential campaign material to strict AI disclosure duties under recommendation n°2026-05, covering AI-generated content in elections, published 7 June 2026. Labels reading "Contenu généré par IA" become compulsory wherever AI substantially shapes campaign material, while faking a declared candidate's likeness without their signed approval becomes illegal from 1 November 2026. Major platforms face a 1 October 2026 deadline to launch tools letting users flag suspect material, feeding a registry that refreshes weekly once the official campaign period begins, giving the public a running record of AI electoral content flagged during the race.
Ålands lagting — Ålands lagting has passed the Landskapslag om tillsyn över artificiell intelligens (AI Oversight Law), transposing the EU AI Act into legislation for the autonomous region. Following two readings, the statute names Datainspektionen as the supervisory authority responsible for enforcement, giving the body powers over AI system oversight and compliance across Åland's jurisdiction. The framework applies AI system compliance regulation modelled on the AI Act's structure, adapted to the islands' devolved legal system.
Ministry of Interior of Serbia — Serbia's Ministry of Interior runs NEC's NeoFace Watch and NtechLab's FindFace, ministry sources told BIRN, extending facial recognition surveillance already built on thousands of Huawei cameras with biometric capability. Russia and Iran use the same software. Rights groups warn this biometric identification system lacks legal grounding, since biometric provisions were left out of the draft Law on Police.
National Agency for the Information Society (AKSHI) — Albania's National Agency for the Information Society has put a first draft of Albania AI regulation to public consultation, running 29 May to 26 June 2026. The proposed draft law on artificial intelligence runs to nine chapters, 106 articles and 13 annexes, closely following EU Regulation 2024/1689 by grading systems on risk, banning manipulative or social-scoring uses outright, and layering heavy technical and governance duties onto anything rated high-risk. AKSHI would coordinate enforcement, and the text bars any legally damaging decision resting solely on real-time biometric identification.
National Agency for Information Society (AKSHI) — The government opened public consultation on its draft Law on Artificial Intelligence, which bases Albania AI regulation largely on the EU AI Act. The draft sorts AI systems by risk, bans manipulative practices and social scoring, mandates human oversight of high-risk systems, makes AKSHI coordinating authority and adds a regulatory sandbox. The AI rules allow real-time biometric identification in public spaces only with a court order for serious crimes.
Ministry of Public Administration of Montenegro — Montenegro AI policy advanced on 27 May 2026 when a Working Group under the Ministry of Public Administration, aided by UNDP, held its fifth meeting to pore over the strategy's initial blueprint for 2026-2030. Participants debated ambitions at both strategic and operational levels, plus yardsticks for gauging success. Attention centred on building digital infrastructure, sharpening skills, and shaping regulatory frameworks — pillars of the emerging national AI framework.
Liechtensteinische Landesverwaltung (LLV) — LLV, Liechtenstein's State Administration, broadened its AI chatbot government website tool LiBot to cover all of www.llv.li, moving beyond its earlier commercial-register-only scope. The assistant, an AI-driven virtual assistant for public sector services, resolves common queries at any hour and steers visitors toward the right information, built on an AI model tuned for clear, accurate replies. Officials frame it as part of a wider digitalisation strategy.
Government of Spain, Ministry for Digital Transformation and Public Service — The Council of Ministers forwarded the draft Organic Law for the Good Use and Governance of Artificial Intelligence to the Congress of Deputies, aligning Spain AI regulation with the EU AI Regulation. Penalties climb to 35 million euros or 7% of global turnover; AESIA, AEPD and CGPJ oversee compliance; a national AI sandbox, a public-body AI catalogue and a public-sector AI delegate follow. By September 2026 this AI legislation remained before parliament, unpassed.
Government of the Republic of Slovenia — A newly formed AI ethics council, the National Council for Ethics in Artificial Intelligence, now has its full roster of members named by the Government of the Republic of Slovenia, acting under authority granted by the ZIUDHPUI law. This independent panel of specialists will judge how artificial intelligence affects society, democracy and ethics, offering guidance to shape responsible public-sector adoption. Beyond issuing formal advice, the body carries a mandate to weigh the legal, environmental and social consequences of AI tools in use, functioning effectively as an ethical AI oversight body for the state.
Upravna akademija (Administrative Academy) of the Ministry of Public Administration — Over 40,000 civil servants in Slovenia completed AI training for civil servants under a scheme run by the Administrative Academy within the Ministry of Public Administration. More than 80 modules spanned generative tools for senior staff alongside routine departmental applications. The Republic of Slovenia and the European Union's NextGenerationEU recovery mechanism co-funded the effort, spreading AI literacy among public sector employees.
Isle of Man Government — Isle of Man AI regulation lets businesses own AI training data as lawful assets after Data Asset Foundations legislation gained Royal Assent, the world's first statutory framework governing data assets. That permits pricing, licensing and safe exchange with collaborators. Follow-on rules and a world-first statutory Data Asset Register are being drafted; Enterprise Minister Tim Johnston argues such AI rules put the island ahead in the worldwide data economy.
Digital Learning Hub and Luxembourg Ministry of State — Luxembourg's Elements of AI literacy training has reached over 7,600 residents since 2021, Minister Elisabeth Margue said, closing its 5th edition. That's roughly 1.1% of working-age adults, beating the 1% target for 2030 ahead of schedule. The free, nationwide AI literacy programme keeps growing.
Town of Otrokovice — Otrokovice, a Czech town, has switched on an AI assistant for council websites so residents can find things faster. Tucked behind a corner icon, the tool takes questions without sending anyone away from the page, combing municipal pages and other material for dependable, up-to-date replies on services, official business and what is happening locally. A spin-off of the Faculty of Mechatronics at the Technical University of Liberec supplies the municipal information chatbot, which the Constitutional Court also runs.
Holy See — Pope Leo XIV has instituted an Interdicasterial Commission on Artificial Intelligence, the forum for Holy See AI policy. Cardinal Michael Czerny, prefect of the Dicastery for Promoting Integral Human Development, signed the rescript, released by the Holy See Press Office; papal assent came at a 3 May 2026 audience. Seven bodies take part: that one, those for the Doctrine of the Faith, Culture and Education and Communication, and the Pontifical Academy for Life, of Sciences and of Social Sciences. Whichever body coordinates — a renewable yearly turn, Czerny's first — must ease cooperation and exchange on artificial-intelligence work, Vatican AI governance rules for use inside the Holy See included.
Government of Malta — Malta became the site of OpenAI's debut national government tie-up, letting citizens and residents on its digital identity registry claim a full year of free ChatGPT Plus, the premium chatbot subscription. Unlocking the offer means finishing a no-cost AI literacy course co-designed with the University of Malta. The Malta Digital Innovation Authority administers the initiative.
Dicastery for Promoting Integral Human Development, Holy See — The Holy See has launched an Interdicasterial Commission on Artificial Intelligence (ICAI) for Vatican AI governance church-wide. Cardinal Michael Czerny, Dicastery for Promoting Integral Human Development Prefect, signed the founding decree after Pope Leo XIV's approval on 3 May 2026. Drawing on seven dicasteries and pontifical academies, the body will pool knowledge and set shared AI policy for the central administration.
Communications Regulatory Agency of Bosnia and Herzegovina — More than 70 percent of staff at the Communications Regulatory Agency of Bosnia and Herzegovina reach for artificial intelligence tools regularly or occasionally, an internal poll found. AI adoption in public administration outpaces the rulebook: the regulator is still drafting one, and no statute covers such work. The Ministry of Foreign Affairs, the Central Election Commission and the Public Procurement Agency, whose chatbot fields queries on its eNabavke portal, report the same, marking AI use across state bodies.
Government of Malta — Malta's Government wants machines handling court sittings, parliamentary sessions and public submissions: a tendered AI speech-to-text system, priced near €147,750, targets 85% accuracy in year one while managing code-switching between Maltese and English. Staff will check and fix computer drafts via a smart editor — an automatic transcription tool for courts and parliament.
Government Digital Service — The Government Digital Service has released GOV.UK Chat inside the GOV.UK app, letting residents pose everyday questions and get replies drawn from verified public-sector sources. This AI government chatbot ran as a soft launch since March, processing over 15,000 queries from more than 7,800 distinct users. It now moves to full public availability, an AI tool for public information queries built into the app.
High Court in Prague — Prague's High Court has handed citizens Čenda, an AI legal guidance chatbot running the Claude model on the Lovable platform. Čenda quotes no statute, sketching what a person divorcing or mired in debt might try, sourced from justice.cz and gov.cz. Issued alongside the automated legal help tool were 18,121 civil and 6,314 insolvency judgments stretching to 2020, the pair billed 800,000 crowns.
Tax Administration of Serbia — Tax inspectors in Serbia spent four days learning AI audit tools for tax administration at a European Union-backed session held 12-15 May 2026 under the IPA 2021 scheme. Markus Drost, the EU specialist running the course, showed how algorithms can surface high-risk audit candidates, though he insisted a person must sign off on every final assessment. Real case studies formed the basis of practical audit simulations, giving officials a chance to test AI-driven risk detection in tax administration hands-on.
Government of North Macedonia — North Macedonia became the 46th signatory to the Council of Europe Framework Convention on Artificial Intelligence, the first legally binding international treaty on this technology. Justice Minister Igor Filkov and Secretary General Alain Berset signed it in Skopje, marking North Macedonia AI regulation progress meant to keep artificial intelligence systems accountable to democratic norms and legal safeguards, part of a wider AI governance treaty push.
Agencija za zaštitu osobnih podataka (AZOP) — AI data protection now has a formal reference point in Croatia: on 7 May 2026, the Personal Data Protection Agency (AZOP) issued guidance covering anyone who builds, trains, tests, integrates or runs AI systems. Titled "Data Protection and Artificial Intelligence", it clarifies GDPR applicability to AI models, assigns controller and processor responsibility, and details lawful processing grounds, privacy-by-design duties, transparency requirements, DPIA thresholds and security measures, alongside model anonymity, LLM risk factors and privacy-enhancing technologies. Organisations gain one benchmark for data protection and AI compliance across the development lifecycle.
Ministry of Economic Development and Digitalization — The Ministry of Economic Development and Digitalization and the Ministry of Education and Research formed the Sub-Council for Artificial Intelligence and Applied Research (SCIA), an advisory body for Moldova AI policy. Joining academia, public authorities and the private sector, SCIA steers EU AI Act transposition into domestic law and names an authority to implement national AI governance.
UK Parliament — The Crime and Policing Act 2026, now law after Royal Assent, lets the Secretary of State rewrite the Online Safety Act 2023 to close AI regulation gaps around illegal AI-generated content and services enabling priority offences. Producing fake nude or sexualised images with AI tools becomes its own crime under this AI image generation law.
Ministry of Public Administration of Montenegro — Montenegro's Ministry of Public Administration has rolled out an AI chatbot for public services, live on the Government and eUprava portals, sparing citizens paperwork to find information. Minister Maraš Dukaj called round-the-clock, 24/7 access the key benefit of this AI-powered government assistant. Still in early pilot stage, it will gain new capabilities as feedback arrives.
Government Office of the Czech Republic — Nine ministries sit among 50 public bodies across Czechia now signed onto the AI National Skilling Plan's 'AI na úřad' scheme, a collaboration led by the Government Office of the Czech Republic with Microsoft and Digiskills to build AI skills training for civil servants. First-year totals put ready-made learning environments at 24,000 workers, while 577 staff inside the Government Office itself wrapped up 45,153 online courses between them. The initiative pairs safe handling of Copilot Chat and comparable workplace software with wider AI literacy for government employees.
Ministry of Economic Development and Digitalization — Public feedback is being gathered in Moldova on a draft National Programme on Artificial Intelligence and Data Management, covering the years 2026 to 2030 and led by the Ministry of Economic Development and Digitalization. The Moldova AI policy seeks to turn the country from a mere buyer of foreign technology into a builder of homegrown AI products, bringing data oversight rules closer to EU standards along the way. This broader national AI strategy signals an effort to grow domestic tech capacity rather than depend on imports.
Comunidad de Madrid — SOL, a new AI public services chatbot from Madrid's Consejería de Digitalización, now serves 7 million residents around the clock. Built on Microsoft Foundry with Azure OpenAI Service, it answers voice and text queries in Spanish, English and Romanian, using over 70,000 official URLs from the regional government and transport authority. Atos, Bravent and T-Systems helped build the digital assistant.
Ministry of Interior and Administration of Poland — Poland's Ministry of Interior and Administration has created a Team for the Application of Artificial Intelligence Systems (Zespół AI), spanning Police, State Fire Service, Border Guard, State Protection Service and Office for Foreigners. Its brief for AI in national security covers biometric identification, emergency-call categorisation, critical infrastructure prediction and large language model applications, plus AI for border control and migration management. A full strategy is due by 2035.
Government of Ukraine — Ukraine's Ministry of Digital Transformation is trialling AI policy impact analysis, software that predicts how draft acts will shape outcomes for citizens, businesses and other groups, labelling each positive, neutral or negative. Acting minister Oleksandr Bornyakov described it as an early-stage experiment, aimed at handing officials quicker, evidence-grounded judgement than manual review offers. This AI-driven policy assessment lacks a set launch date beyond the current pilot.
Dutch government — Ten sitting watchdogs, not one new body, will handle Netherlands AI Act implementing legislation under the Uitvoeringswet verordening artificiële intelligentie, open for consultation. These include the Dutch Data Protection Authority (AP) and the Authority for Digital Infrastructure (RDI), which also jointly steer a cross-sector AI regulatory sandbox proposed in the same bill.
UK Parliament — New rules under the Data Protection Act 2018 require the Information Commissioner to produce a statutory Code of Practice on AI regulation compliance, addressing how organisations handle personal data while building and deploying Artificial Intelligence and automated decision-making tools, once regulations took effect on 12 May 2026. Special attention must go to safeguarding children's personal data within these automated systems. The outcome: formal UK data protection guidance on AI use.
Ministry of Economic Development and Digitalization — Moldova's Ministry of Economic Development and Digitalization and the e-Government Agency have drafted a bill legalising AI in public services from 1 January 2027, provided humans oversee outcomes. Fully automated rulings stay barred for sensitive cases like suspending welfare payments or settling child custody, though citizens gain appeal rights. This tightens safeguards on AI in government administration.
Primăria Sectorului 3, Bucharest — Romania's Primăria Sectorului 3, Bucharest's Sector 3 city hall, now answers residents' calls with an AI phone assistant handling routine administrative queries instead of a human operator. Reported on 21 April 2026 after weeks of use, the intelligent virtual assistant was still in calibration; the city hall admits errors can occur meanwhile.
Regierung des Fürstentums Liechtenstein (Government of the Principality of Liechtenstein) — Public-sector AI use in Liechtenstein now runs on four pillars: accountability, putting people first, openness, and innovation-driven efficiency. The recently approved KI-Strategie der Liechtensteinischen Landesverwaltung (AI Strategy of the Liechtenstein State Administration) marks an AI strategy for public administration, structured around four action fields: people, organisation, governance, and infrastructure, aimed at steering AI adoption in government agencies. Officials set no fixed end date, but a fundamental revision must happen no later than 2030.
Financial Administration of the Czech Republic — The Ministry of Finance and the Financial Administration have opened a beta of an AI tax return filing tool, which the ministry calls a first for Czech public administration. Upload a payslip and the software picks out the figures required, dropping each into the right box, so nothing needs retyping and no row numbers need finding. Automated tax form completion covers only wage earners for now, including people paid by several employers and students earning on the side.
Municipality of Ratiboř — Ratiboř, a village in the Zlín Region, now fields enquiries through an AI mayor — a digital twin of Martin Žabčík, built over two years by Digiregion with AI Produkce. The likeness appears on a video screen in the council building's entrance hall, hooked into the village's own knowledge store, so callers learn of roadworks or planned spending. Because the digital council avatar talks day and night, residents wary of laptops and mobiles gain too.
Government of the Slovak Republic — Slovakia has created the Rada vlády SR pre digitálnu transformáciu a umelú inteligenciu (Government Council for Digital Transformation and AI), on a MIRRI proposal. It replaces the former Government Council for Digitalisation of Public Administration and the Single Digital Market, steering Slovak AI policy and digital transformation through coordinating work spanning AI governance coordination government-wide.
Občina Sveta Trojica v Slovenskih goricah (Municipality of Sveta Trojica) — Klara, an AI digital secretary, has fielded calls for Sveta Trojica municipality since 2026. Unveiled by Mayor David Klobasa on Easter Sunday and live from that Friday, the tool grasps what callers want, replies, and auto-drafts written summaries as official records. Council goals cited included speed, order and openness, though human contact, officials insisted, stays central: artificial intelligence for municipal call handling will not replace it.
Agencia Estatal de Administración Tributaria (AEAT) — Spain's AEAT tax authority deploys AI forensic analysis tax agency tools, rolling out 500 licences of Cellebrite Pathfinder to sift unstructured digital evidence—emails, images, videos—in open cases. The tool flags patterns and anomalies in seized data without replacing human judgement, an AI-powered digital evidence review layered atop inspectors' work. AEAT paid roughly €326,000 for the software in 2023; fiscal law experts warn its reach may extend beyond verifying income to mapping a taxpayer's social and family circle.
Government of Estonia — Ministers approved and passed to parliament a bill setting a technology-neutral legal frame around automatic administrative procedure — Estonia AI regulation reaching routine, repeated and data-driven rulings. Central and local bodies would gain authority to decide through algorithms and machine learning, weighed against citizens' constitutional guarantees. Where an automated route proves opaque or ill-suited, the applicant may turn to a person, a safeguard inside the AI rules for administrative decisions.
Rada Ministrów (Council of Ministers of Poland) — AI regulation Poland now covers content labelling, a public registry for sanctioned or withdrawn systems, and legal grounds for AI regulatory sandboxes, following approval by Poland's Council of Ministers. Officials call it the EU's first such wide-ranging AI policy framework, with the draft AI Systems Act next heading to parliament.
Greek Ministry of Digital Governance and AI — Greece's Ministry of Digital Governance switched on pilot AI traffic cameras at eight high-risk points in Attica on 28 March 2026, certifying violations through the traffic police system and notifying drivers via gov.gr. The first 130 digital fines, mostly for red-light running and helmetless motorcyclists, went out within days; three ministries share the programme: digital governance, citizen protection and transport.
Gibraltar Parliament — Gibraltar Parliament has opened the whole legislative archive — every Hansard volume back to 1969 — through a new AI Hansard search built for public use. Beyond debate transcripts the service reaches Speakers' Rulings and Standing Orders, turning precedent hunting that once required manual research into a single query. This AI-assisted parliamentary research assistant condenses debates neutrally, yet cannot draft communications, speeches or anything original. The interface is already running on parliament.gi.
Canton Ticino — Ticino's public administration has rolled out an AI translation for government staff tool built on Switzerland's open-source Apertus language model, with roughly 100 civil servants trialling it in the initial rollout phase. Artificialy, the Ticino-based firm behind the system, designed it to convert text across German, French, Italian, Romanian and Ukrainian while keeping sensitive data within Swiss borders, a multilingual public-sector translation approach that could inform wider cantonal adoption.
Supreme Court of Montenegro and Judicial Council — Montenegro's Supreme Court and Judicial Council have rolled out an AI case law system built in-house by the Council's ICT department. It gives judges intelligent search across rulings plus an AI chatbot, unveiled at the Supreme Court on 8 April 2026, functioning as an AI-powered judicial search platform for the courts. A trial is under way at the Commercial Court, with a wider rollout planned across all Montenegrin courts.
Government of Luxembourg — Luxembourg's Council of Government now requires human review of AI in central administration before content is published, holding the approving civil servant accountable rather than the system itself. Disclosure duties also apply, matching Article 50 of the EU AI Act, part of wider AI governance in public administration.
Austrian federal government (State Secretary for Digitalisation) — Austria's State Secretary for Digitalisation, Alexander Pröll, says the federal electronic file system (ELAK) will add AI in the second half of 2026, letting 13,000 staff query whole file collections and get structured AI document summaries instantly. It is among five Public AI initiative applications, alongside KI-SUN, a knowledge assistant already covering 6,000-plus pages of federal training material.
Finanzamt Hilden — Employee tax cases in North Rhine-Westphalia now get AI tax risk management, with Finanzamt Hilden among the offices first to run it statewide in Germany. Risk-oriented selection flags audit-worthy returns, sorts low-risk filings, and clears some 2025 returns via AI-based tax audit selection with no human review. Eight offices piloted the approach from May 2025 before the wider launch.
Ministry of Digital Transformation — Slovenia's Ministry of Digital Transformation has adopted the Guidelines for the Development, Introduction and Use of Artificial Intelligence in Public Administration, a framework meant to steer government bodies toward deploying AI in public administration safely, ethically and effectively when serving citizens and businesses. Rather than binding rules, officials describe it as offering practical grounding for weaving automated systems into routine state functions — part of a broader push toward artificial intelligence in government services. Two attached annexes fill in specifics: guidance for staff on handling publicly available generative tools, plus an exit plan for cloud providers spelling out mandatory security benchmarks.
Национална агенция за приходите (NRA) — Bulgaria's National Revenue Agency is extending AI in tax administration to all its units after a successful pilot, the first Bulgarian administration to use the national BgGPT model at scale. Built with INSAIT, it covers head office and seven territorial directorates, targeting everyday use by 7,000-plus staff searching specific cases under shifting tax and social-security law.
CEPOL and EU4Security Moldova — Over 20 Moldovan officials finished a five-day AI investigations training course in Chișinău, run by CEPOL under the EU4Security Moldova project. Lessons spanned basics, data management, software tools, spotting patterns, and whether AI-derived evidence holds up in court, plus ethics and rights safeguards. It builds Moldova's capacity for AI in criminal investigations.
Government of Romania — Four separate bodies, not a single regulator, will police EU AI Act Romania compliance under a Memorandum passed by the country's cabinet. ANCOM leads as the designated contact point and general market watchdog, ASF and BNR govern high-risk AI in financial services, ANSPDCP handles biometric technology and law-enforcement applications, and ADR vets conformity assessment bodies. Regulation 2024/1689 thus takes effect through this layered domestic structure rather than a unified agency.
Government of the Republic of Slovenia — Five pillars anchor the national AI strategy of Slovenia, set by the Government of the Republic of Slovenia as NsUI 2030 on 5 March 2026 for the period to 2030: sovereign, trustworthy AI; broader uptake across business and government; domestic AI skills and ecosystem growth; defence and security uses; and stronger global ties. This Slovenia AI policy supersedes the earlier National Programme for AI by 2025.
Government of Luxembourg — Luxembourg's government, launching the AI4LUX campaign, said every civil servant would get a sovereign AI chatbot within weeks, hosted on site under its Mistral AI partnership so data stays in the country. Staff are to process sensitive information securely with it and create personal agents for routine work. A legal chatbot for the Legilux platform is in development, and one for the most-visited state websites is being prepared.
Informacijski pooblaščenec (Information Commissioner of the Republic of Slovenia) — Slovenia's Information Commissioner issued guidance on responsible AI use in government, covering when personal data may enter AI tools and when an individual counts as non-identifiable. It requires AI answers to be verified and confirms decisions stay with a human official, tying safe AI adoption in the public sector to data protection rather than the tools themselves.
Government of Moldova — Moldova has paused Moldovan AI law drafting on EU advice, said Michelle Iliev, State Secretary at the Ministry of Economic Development and Digitalization. The European Commission is simplifying its own AI Act via the Digital Omnibus package, making a parallel effort unnecessary. AI regulation resumes once the EU framework is finalised.
Constitutional Court of the Czech Republic — Czechia's Constitutional Court has switched on a rebuilt virtual assistant for AI case-law search on the court's pages and in NALUS, after weeks of testing. Liberec's Technical University spin-off nanotrix.ai wrote the system, described by the Court as the earliest such tool on any Czech court site or judgment archive. The judicial database chatbot fields spoken queries and queries in other languages; users must verify whatever comes back.
Gibraltar Regulatory Authority — The Gibraltar Regulatory Authority, acting as Information Commissioner, is among 61 signatories to a Joint Statement on AI-Generated Imagery, convened by the Global Privacy Assembly. Signing sets Gibraltar AI regulation alongside a common line on artificial intelligence that fabricates convincing photographs and video of identifiable people who never consented, children most at risk. Builders and users must install AI imagery safeguards, be candid about capability, pull harmful material on request, and shield children.
Office of the Data Protection Authority — Guernsey's Office of the Data Protection Authority is among 61 regulators behind the Joint Statement on AI-Generated Imagery and the Protection of Privacy, which the ODPA convened as secretariat and co-chair of the Global Privacy Assembly's International Enforcement Cooperation Working Group. Guernsey AI regulation now bites on systems fabricating lifelike pictures and video of real people who never agreed, minors and vulnerable groups most at risk, and co-signatories want strong safeguards, candour about system capability, reachable takedown routes for damaging material, and added protection for children. Parallel AI image rules form a Crown Dependency Data Protection Advisory on generative AI image creation, co-issued with the Isle of Man Information Commissioner and Jersey's Office of the Information Commissioner, restating that indecent images of a child stay criminal, AI-made or not, compliance resting with whoever builds and uses them. Islanders are told to tighten social-media privacy settings, treat AI apps wanting photographs or personal data warily, grasp their entitlements, and discuss internet safety with children.
Gibraltar Regulatory Authority — The Gibraltar Regulatory Authority co-signed a Joint Statement on AI-Generated Imagery with 60 other data protection authorities on 23 February 2026, addressing AI image generation privacy risks from realistic images and videos of identifiable individuals made without consent. It urges safeguards, transparency, removal mechanisms, and protections for children against AI-generated imagery of identifiable individuals.
Nemzeti Adó- és Vámhivatal (NAV) — Hungary's Nemzeti Adó- és Vámhivatal now grounds its 2026 audit programme in AI tax audit selection. Vágujhelyi Ferenc, who leads NAV, sketched a single sweep across invoice feeds, till records, eKÁER consignment filings, sectoral declarations and cross-border money-flow reporting. Risky traders surface the instant a bill is cut, letting AI-driven risk screening put a reviser onto a fraudster ahead of any filing.
Regeringskansliet (Finansdepartementet) — Sweden's new AI policy, Sveriges AI-strategi, supersedes the 2018 National Approach. It targets a top-ten global ranking for artificial intelligence and world-leading AI adoption for public administration.
Greek Ministry of Interior — Greece's Ministry of Interior renewed its Google cooperation deal, extending AI training for civil servants through December 2026 with Gemini AI built into the curriculum. Since 2023, 85,772 staff have finished digital skills courses, including 35,839 who joined AI-focused sessions across 77 webinars in 2024-2025, averaging 1,130 attendees each. The new phase targets hands-on AI fluency in daily public-sector work, broadening earlier government employee AI skills efforts.
Belgeodesiya — Belgeodesiya is putting AI building recognition into the digital foundation of the State Navigation Map of the Republic of Belarus, per the State Property Committee. Accuracy runs from 65 to 90 per cent, varying with how many storeys a structure has. Settlements across 40 administrative districts have gone through satellite-image building detection, thickening the map's Buildings layer.
Administrative Court of Tirana — Diella, the AI minister Albania built from Anila Bisha's voice and image, is now the subject of a court fight the actor launched to halt the government's use of her likeness. Bisha maintains her signed deal covered only a digital helper on e-Albania, not a cabinet post, and that officials skipped her consent before the switch. Lawyer Aranit Roshi labels the injunction the first legal step against the alleged misuse, with a further damages claim pending.
Estonian courts, prosecution and local councils — Machine-drafted filings now reach Estonian courts, prosecutors and councils often enough that officials lose days reading them and verifying citations that frequently prove false — the visible cost of AI in public administration. Tartu Circuit Court docked a state-appointed defence lawyer's fee by two thirds in February after an appeal leaned on Supreme Court judgments that either meant something else or were wholly fictitious; Chief Prosecutor Astrid Asi says applications of 50 or 60 pages are no longer unusual, since bulk and footnotes cost nothing to produce. Mayors describe the same drag: Põhja-Pärnumaa's Madis Koit watched hundreds of generated questions arrive during wind farm planning a year ago and freeze other business, while Rae's Gerli Lehe has clerks combing each letter line by line. A prosecution test group is meanwhile exploring generative AI in government, with Asi insisting a human makes the closing call.
Hungarian Police — Misidentifications since January have dented Hungarian police facial recognition, used even for petty offences. A Nyíregyháza secondary school pupil was pursued as a fraud suspect while sitting in class 200 kilometres away; a Szeged woman stood accused of lifting a plush bear in Budapest; a 19-year-old was pinned to a shampoo theft. All three had alibis. Who supplied the kit is unpublished; Remport Ádám suspects the biometric identification cameras are Chinese-made.
Ministry of Justice of North Macedonia — North Macedonia's Ministry of Justice opened consultation until mid-March on draft changes to the Law on Personal Data Protection, tackling North Macedonia AI regulation for the first time. Filed via the ENER platform in February 2026, the text lets firms cite legitimate interest for machine learning systems personal data processing during training and testing, given safeguards, plus strict data-minimisation caps.
Bundesakademie für öffentliche Verwaltung (BAköV) — The Federal Academy of Public Administration (BAköV) issued AI learning pathway (KI-Lernpfad) version 2.1, offering AI training for civil servants. Built with the AI Advisory Centre (BeKI), the pathway covers AI literacy basics, AI Act data protection and role-specific modules. Every German federal employee, not only AI or digitalisation officers, can join the civil service AI education.
Autoriteit Persoonsgegevens (AP) — The Autoriteit Persoonsgegevens (AP), the Dutch Data Protection Authority, finalised its stance on Netherlands AI regulation for generative AI systems, setting conditions for safe, responsible use that respects fundamental rights. A 'values at work' scenario weighs innovation against safeguards for democracy and civil liberties. The guidance sketches future paths for generative artificial intelligence oversight rather than fixed rules.
Rijksorganisatie voor Ontwikkeling, Digitalisering en Innovatie (RADIO) — Dutch civil servants must now finish AI training for civil servants, a 30-minute e-learning module capped by a quiz, built into the Basismodule Digitaal vakmanschap (Basic Digital Skills) programme. Content spans what AI actually is, situations where employees might run into it, and expectations for handling it responsibly across government. Officials will absorb this AI training for government workers into the compulsory Ambtelijk Vakmanschap (Civil Service Professionalism) curriculum going forward.
HM Government of Gibraltar — Gibraltar's Central Arrears Unit now runs on a bespoke AI debt collection database, built after a hiring round for several vacancies yielded a single suitable applicant. Nigel Feetham, Minister for Justice, Trade and Industry, explained the gap; IT Development Manager Karon Cano wrote the model in-house with Steven Fernandez. Machine-assisted arrears tracking replaces hand-keyed spreadsheet extracts and saves one officer at least an hour's work. A chatbot with no outside feeds returns priority-debt listings.
General Council of the Judiciary (CGPJ) — Spain's General Council of the Judiciary (CGPJ) approved Instruction 2/2026 on 28 January 2026, setting AI rules for judges and magistrates, generative AI included. Judges may use only AI systems from the competent justice administrations or the CGPJ, and AI may never replace them in deciding cases. The Official State Gazette (BOE) then published it.
Consejo General del Poder Judicial (CGPJ) — Spanish courts face binding curbs on AI use by judges under Instruction 2/2026 from the General Council of the Judiciary (CGPJ), published in the BOE on 30 January 2026. Algorithmic tools must stay under human command, autonomous rulings are barred, and only CGPJ-vetted software may be deployed. Profiling, behavioural forecasting and specially protected personal data via judicial artificial intelligence tools stay off-limits.
Government Office (Riigikantselei) — Estonia formed the Eesti.ai Advisory Board to steer AI in government under Bolt founder Markus Villig, who chairs the panel. Answering straight to Prime Minister Kristen Michal, it supervises major artificial intelligence work spanning the whole economy, alongside other prominent Estonian tech entrepreneurs and experts.
Department of Foreign Affairs and Trade — Ireland's Department of Foreign Affairs and Trade is piloting Ask Finn, an internal AI chatbot for passport processing, as Passport Service, Revenue and Citizens Information Board officials prepared for the Oireachtas Committee on Artificial Intelligence. Revenue's AI models route taxpayer queries more precisely to caseworkers, and Citizens Information is rolling out its AskCI bot.
Consell Superior de la Justícia — Andorra's Consell Superior de la Justícia has switched on an AI case-law search holding 15,000-plus rulings, open to lawyers and citizens. Machine-generated metadata tags each entry, letting queries run in plain speech or boolean operators across every decision under one statute. A 4,000-term thesaurus guides the rulings archive, with ECLI codes for cross-border traceability.
Isle of Man Government, Department for Enterprise (Digital Isle of Man) — £1 million backs a freshly created National AI Office (NAIO) steering Isle of Man AI regulation, overseeing how artificial intelligence takes hold across the Island. NAIO merges scattered government and industry roles under one roof, giving guidance aimed at growth, better public services and stronger AI literacy. Built atop Activate AI, the 2024 programme, it hands the Island's artificial intelligence strategy a lasting institutional home.
Academy of Management under the President of the Republic of Belarus — Civil servants across Belarus can now take a one-day generative AI training for civil servants at the Academy of Management under the President of the Republic of Belarus. The short course builds skill with ChatGPT, YandexGPT, Gemini and similar systems for producing text, images, audio and video, adding prompt techniques suited to state-body work. It reflects wider momentum behind AI adoption in public administration, with sign-ups running from January to July 2026.
Parliament of the Czech Republic — Czechia's Criminal Code (Act No. 40/2009 Sb.) now covers AI deepfake criminalisation under new Section 191a, punishing non-consensual pornographic deepfakes made or shared without the depicted person's agreement. Offenders face up to two years in prison for a standard breach, rising to five years for aggravated cases.
National Council of the Slovak Republic — Slovakia's Act No. 318/2025 Coll., a Slovak AI regulation, folds the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) into domestic law, amending Act No. 56/2018 Z.z. on product conformity assessment. High-risk AI systems face new national conformity procedures, duties for economic operators, and enforcement mechanisms with penalties. AI system compliance follows.
The EU AI Act's high-risk AI rules for Annex III systems, covering migration, border control and law enforcement, missed their original 2 August 2026 start. The Digital Omnibus on AI, Regulation (EU) 2026/1744, moved them to 2 December 2027, taking effect on 27 July 2026, six days before the original deadline.
The UK Home Office launched PoliceAI, a national centre for AI in policing, on 10 June 2026 with £75 million over three years to find, test and scale AI tools across the 43 forces of England and Wales. Named uses: facial recognition, deepfake detection, call transcription, triage of non-emergency calls and digital evidence, disclosure and audio-visual redaction. It says the tools could eventually free the equivalent of 3,000 officers.
Essex Police suspended live facial recognition, citing accuracy and bias risks, once it emerged its match threshold had been calibrated for another vendor's algorithm. The UK Information Commissioner's Office then audited six forces and made 107 recommendations, 93 fully and 14 partly accepted, finding no standard practice of routine bias testing.
The UK government's TechFirst programme commits £187 million to digital and AI skills training: TechYouth (£24m, 1 million secondary students over three years), TechGrad (£96.8m, 1,000 undergraduate scholarships a year, 100 Research MSc places, 100 elite AI scholarships), TechExpert (£48.4m, 500 domestic PhD students) and TechLocal (£18m, regional seed funding). Separately, industry and government aim to equip 7.5 million workers, roughly 20% of the workforce, with essential AI skills by 2030.
The EU's Digital Europe Programme has put over €294 million into digital skills funding for upskilling, reskilling and skilling the workforce. An EU-funded AI Skills Academy launches in 2026.
Hellenic Data Protection Authority (HDPA) — Greece's HDPA outlawed the Hellenic Police's €4 million Smart Policing scheme under Decision 45/2025, banning AI-enabled surveillance in law enforcement. Scanning faces and fingerprints via handheld field gear, the biometric identification system for policing lacked lawful footing, regulators found, and officers were told to keep it off.
Polish Ministry of Digital Affairs (Ministerstwo Cyfryzacji) — Citizens across Poland can now ask an AI virtual assistant for government services directly inside the mObywatel app, following its 31 December 2025 debut from the Ministry of Digital Affairs. The tool runs on PLLuM, a homegrown Polish Large Language Model, and works as an AI chatbot for public services, walking people through tasks such as requesting a passport or ID card, securing a PESEL number, or notifying authorities of a car sale. It cannot see personal records or tap into state databases, and only reached the public after a two-month trial limited to 5% of users.
Council of Ministers of the Republic of Belarus — Belarus's Council of Ministers has mandated AI in public administration through Resolution No 793, backing surveillance video analytics powered by artificial intelligence and citizen-facing digital assistants with over 1.08 billion Belarusian roubles through 2030. Three strands carry the work: 'Data Economy', 'Digital State' and 'Digital Sovereignty', part of the 'Digital Belarus' programme effective 1 January 2026.
Dún Laoghaire-Rathdown County Council — Gaist's AI road condition survey system now covers 823 km of routes for Dún Laoghaire-Rathdown County Council, Ireland's first such pilot by a local authority. Vehicle cameras capture 360-degree footage that AI software turns into surface ratings, faded-marking alerts and street furniture maps, feeding wider AI-based infrastructure monitoring.
Arbeidstilsynet (Norwegian Labour Inspection Authority) — Arbeidstilsynet, Norway's Labour Inspection Authority, has launched an AI chatbot for foreign workers built on material drawn from arbeidstilsynet.no. The AI-powered assistant spans the regulator's entire area of responsibility and replies in 12 languages: Norwegian, English, Bulgarian, Estonian, French, Latvian, Lithuanian, Polish, Portuguese, Romanian, Russian, Spanish and Ukrainian, helping migrant employees get guidance in their own tongue. Its debut supports a broader state campaign targeting social dumping and workplace crime.
Digitaal Vlaanderen (Flemish Government) — Belgium's Digitaal Vlaanderen built an AI literacy framework using five AI personas, showing local and Flemish civil servants which duties and skills each role needs. UNESCO co-developed the tool, which meets the EU AI Act's public-sector AI competence mandate and points staff toward linked training courses.
Ministry of Investments, Regional Development and Informatization of the Slovak Republic (MIRRI) — A national AI strategy roadmap now guides Slovakia: MIRRI has released the AI Vision for Slovakia (Vízia AI pre Slovensko), mapping the country's AI adoption ambitions. Six themes anchor it — computing infrastructure, data sovereignty with administration, economic growth, ethical protection of fundamental rights, skills-building, and European and global cooperation — an AI policy framework preceding a fuller National AI Strategy due in 2026.
Ministero del Lavoro e delle Politiche Sociali — Italy's Ministry of Labour adopted Decree No 180, setting Guidelines for the Implementation of Artificial Intelligence in the World of Work, giving employers a phased roadmap for AI in the workplace. The guidelines cover artificial intelligence in employment through alignment with the EU AI Act and GDPR, addressing worker privacy, non-discrimination and human oversight.
Basic Court in Zrenjanin (Osnovni sud u Zrenjaninu) — Instant written transcripts now emerge from Zrenjanin's Basic Court hearings through Jurisdikta, an AI speech-to-text court system offering speech recognition for legal proceedings, confirmed live by the presiding judge. Serbia's judicial officials additionally rolled out Transkripta, turning 800 hours of recorded audio into text inside a single 24-hour span, alongside Legislativa, a chatbot pulling answers solely from domestic statutes, regulations and precedent rather than the broader internet. A gathering on Kopaonik mountain hosted the debut of all three systems.
Spanish AI Supervisory Agency (AESIA) — Sixteen guides now anchor Spain's EU AI Act compliance push, issued on 16 December 2025 by the Spanish Agency for the Supervision of Artificial Intelligence (AESIA) to steer organisations through obligations for high-risk systems under the AI Regulation. Built from findings inside the Spanish AI sandbox, the set opens with an introduction to the framework, adds thirteen technical documents tackling risk management alongside data governance, transparency and cybersecurity duties, and closes with a checklist. None of this AI Regulation compliance guidance is legally binding, though each document aligns its advice with what regulators expect.
Hamburg Senate Chancellery — Roughly 40,000 civil servants now use LLMoin, the City of Hamburg's GDPR-compliant AI writing assistant for government use, after a phased rollout. Dataport built it to draft, summarise and search documents. Bremen, Niedersachsen, Sachsen-Anhalt and Rheinland-Pfalz separately adopted the same AI text drafting tool for civil servants.
Instituto de Gestão Financeira e Equipamentos da Justiça (IGFEJ) — IGFEJ (Institute for Financial Management and Justice Equipment) switched on AI call routing for court services within Portugal's justice administration. An audio bot answers callers, works out their needs, and forwards requests automatically instead of a human operator. Coverage at start spans costs, registry, judicial matters and the technology catalogue.
Hungarian Parliament / Government of Hungary — A new fixture in Hungary AI policy: the Hungarian AI Council (Magyar Mesterséges Intelligencia Tanács), formed under Act LXXV of 2025 solely to advise, with no decision power. The AI government commissioner leads it, joining ministries, regulators MNB, NAIH and GVH, and bodies like the Hungarian Academy of Sciences and HUN-REN on Hungarian AI governance.
Ministero del Lavoro e delle Politiche Sociali — Italy's Ministry of Labour has established an Observatory on the Adoption of Artificial Intelligence Systems in the World of Work under Ministerial Decree No 173, aimed at tracking AI workplace adoption nationally. The decree names only the creation of this monitoring body as its purpose, leaving the wider adoption of artificial intelligence systems in workplaces otherwise unaddressed for now.
San Marino12 Dec 2025San Marino Innovation / ULPA (Ufficio per il Lavoro e le Politiche Attive) — Civil servants at the Ufficio per il Lavoro e le Politiche Attive (ULPA) in San Marino launched a public sector AI training programme with San Marino Innovation, under a protocol signed 27 November 2025. Three hours covered large language models, automation tools like N8n, and low-code systems, showing staff how to simplify file handling and internal workflows. Participants also aimed to build software for public employment services, part of a drive toward AI automation for civil servants.
Conseil d'État (Council of State) — France's Conseil d'État binds its administrative judiciary to seven rules on AI in administrative courts, spanning the Council of State, appeal-level courts and the National Court of Asylum. Judges keep final say over rulings, while French staff face new duties on openness and data protection under this judicial AI governance charter.
Department for Work and Pensions — The Department for Work and Pensions now applies AI vulnerability detection in welfare correspondence, running its Whitemail Insights and Vulnerability Scanner across roughly 25,000 letters and scanned documents daily. Built on an LLM, the tool spots warning signs spanning mental health, suicidal thoughts, hardship and abuse at home, generating AI-flagged vulnerable citizens for staff attention. Caseworkers still judge every flag.
Government of Portugal — Portugal's Council of Ministers adopted the National AI Agenda (ANIA): 32 initiatives spanning four pillars — data, innovation, skills and ethics. This Portugal AI policy targets Europe's top ten digital leaders by 2030, funding a Public Administration Centre of Excellence, faster Portuguese civil-servant AI training, plus health and industry hubs. This national AI strategy leaves the 2030 timeline unstated.
TNO — Dutch government AI applications climbed tenfold within twelve months, from just 8 in 2024 to 81 by 2025, per TNO's Overheidsbrede Monitor GenAI for the Ministry of the Interior. Of these 3 categories tracked, municipalities top the adoption list with 34 tools, largely chatbots sorting information, while 37 of 81 (46%) have moved into everyday routine use. For most public bodies, generative AI use in government still means backend support: 78% serve internal workflows rather than resident-facing tasks.
French Government / Ministry of Justice — A newly formed directorate, DPIA, now sits inside the French Ministry of Justice's secretariat-general, tasked with pushing AI in judicial administration forward under Élise Farge Di Maria's leadership. Magistrates and court agents are already trialling an AI assistant built on sovereign hosted infrastructure, handling document analysis, drafting, summarising and translation. Rather than a one-off project, this establishes AI adoption in courts through a permanent internal structure.
Government of Hungary — A fresh institutional structure for enforcing the European Union's AI Act now exists in Hungary, thanks to Act LXXV of 2025 — a landmark piece of Hungarian AI regulation. Oversight duties split between two bodies: the National Accreditation Authority handles notification tasks, while a designated ministry office leads on market surveillance. Backing them is a newly formed advisory group, the Hungarian Artificial Intelligence Council, and violators risk penalties climbing to EUR 35 million or 7% of worldwide annual turnover. This EU AI Act implementation brought most requirements into force.
Dutch Police (Politie) — Unclear results ended a decade of Dutch predictive policing: the Crime Anticipation System (CAS), running since October 2015, has been phased out. A logistic regression model scanned historical police records to flag likely crime times and locations. An internal evaluation cited murky technical performance and doubtful payoff for this crime forecasting system, triggering its retirement.
Urząd Miasta Poznania (Poznań City Hall) — Poznań is piloting ShpaRAG, an AI assistant for citizen services built on Poland's PLLuM language model, developed with the HIVE AI consortium and NASK. It uses retrieval-augmented generation so call centre staff find answers to resident queries faster. As an AI-powered public service assistant, it joins other tools the city runs, including a generative AI helpline consultant and voicebots for appointment reminders.
States of Guernsey Committee for Home Affairs — Guernsey wants AI deepfake pornography made a crime, under reforms to the Sexual Offences Law 2020 spanning nine areas of sexual offending. Ministers also propose banning cyberflashing, requesting fabricated images, and holding indecent cartoons of children. Police raised the issue after meeting cases of non-consensual synthetic images current rules failed to cover.
Berlin Senate Chancellery — Berlin's Senate Chancellery has equipped 40,000 civil servants with BärGPT, a government AI assistant for public administration offering drafting, translating, analysing and summarising via a secure chat interface. CityLAB Berlin built it as open-source software on GDPR-compliant servers in Germany, an AI tool for civil servants made locally, not licensed from a private vendor.
Office of the Data Protection Authority — Anyone in Guernsey handling personal data now has a ten-point checklist for AI in data protection, published by the ODPA. It explains keeping automated systems transparent and accountable, meeting AI governance requirements under the Data Protection (Bailiwick of Guernsey) Law 2017.
Isle of Man Government (Cabinet Office) — 350-plus civil servants in Isle of Man's government now rely daily on Microsoft Copilot under an AI-driven government productivity scheme, the Efficiencies and Change Programme, targeting £50 million saved within five years. Cabinet Office confirmed by May 2025 departments were digitising workflows via AI tools boosting workforce output.
Agenția Națională pentru Ocuparea Forței de Muncă (ANOFM) — Moldova's National Employment Agency (ANOFM) has fitted its angajat.md platform with an AI chatbot for job seekers, built with the Ministry of Labour and Social Protection and Norwegian government backing. It speaks Romanian, Russian, Ukrainian and English, covering vacancies, unemployment registration, training and support schemes. This AI-driven employment service aims to lift staff workload by handling repeat questions.
Integritetsskyddsmyndigheten (IMY) — Sweden's IMY has ended its inquiry into Försäkringskassan's AI welfare system, closing the case after the agency pulled the machine learning-based risk profiling tool it used to flag temporary parental benefit (VAB) claims for added scrutiny. Media coverage prompted IMY to open the investigation in June 2025; Försäkringskassan withdrew the tool soon after and ruled out bringing it back, so IMY judged the danger had passed.
FITKO / Behördennummer 115 — Berlin, Frankfurt am Main, Aachen, Essen, Magdeburg, Harburg, Königstein and Grünheide are the first eight municipalities running an AI chatbot in government services, launched by Behördennummer 115, Germany's central citizen-service phone line. Föderale IT-Kooperation (FITKO) built the assistant on a vendor-neutral open-source platform powered by gpt-4.1-mini, letting residents get round-the-clock answers to administrative questions directly through their local council websites. The pilot signals an early German push toward automated public-sector support, with the participating cities serving as test cases before any wider network expansion.
Ministry of Digital Affairs — Poland's Ministry of Digital Affairs revised its Policy for the Development of Artificial Intelligence in Poland until 2030 on 17 November 2025, adding indicators to gauge national AI strategy Poland progress. Łódź and the GZM Metropolis join leading centres from 2027; local government pilots PLLuM, and AI HUB Poland readies rollout. An AI adoption roadmap for key industries completes the update.
Stadt Essen — A pilot scheme running in Essen puts a new AI chatbot for citizen services to the test, making the city one of Germany's earliest adopters within the 115-Verbund network from 17 November 2025 onward. Answers on local government matters come from two sources feeding the AI-based municipal service assistant: the established 115 knowledge database and handpicked pages from the council's own site. Berlin, Frankfurt, Magdeburg and Aachen share in the two-month trial, which may extend to every 115-Verbund participant once finished.
Personal Data Protection Agency of Bosnia and Herzegovina — Bosnia and Herzegovina's Personal Data Protection Agency now flags which processing activities demand an AI data protection impact assessment, targeting automated and algorithmic decision tools. Triggers include extensive profiling, decisions made without human input, marketing or profiling of children's records, bulk handling of sensitive categories, biometric and genetic identifiers, and public-space surveillance. Staff monitoring software and background-check record matching complete the AI risk governance obligations.
Ministry of Finance of the Russian Federation — AI budget application verification has arrived at Russia's Ministry of Finance: agents built on Sber's GigaChat and Yandex's language models screen spending requests against state programmes and national projects. A review once taking three to four hours now finishes in just over a minute; at peak volume, up to 450 submissions daily, a multi-day backlog clears within one. Workload has dropped by as much as 70%, part of a wider push toward automated government spending review.
Bulgarian National Audit Office — Odis, a new tool for AI procurement audits from Bulgaria's National Audit Office and INSAIT, Sofia University's AI institute, shrinks a four-day review to one day. Powered by the BgGPT model, this AI assistant for procurement auditors trawls the Electronic Public Procurement Platform, extracting and condensing records to flag issues faster. The Prime Minister attended its debut.
Gouvernement français (Conseil des ministres) — The CNIL would lead market surveillance of high-risk Annex III systems plus prohibited practices under article 24 of the DDADUE bill, which names around fifteen national competent authorities, charged with policing the EU AI Act, after going before the Council of Ministers. In this France AI regulation, Arcom oversees media and transparency obligations while the DGCCRF coordinates as single point of contact. The Senate approved an amended text on 17 February 2026; the National Assembly still has to examine the AI rulebook.
Consejo General del Poder Judicial (CGPJ) — Spain's General Council of the Judiciary (CGPJ) has released KENDOJ, an AI in judicial administration package built by the Centro de Documentación Judicial (CENDOJ), for nationwide court use. This AI toolkit for judges auto-drafts case summaries and offers a chat-style tool for real-time legal queries. CGPJ states the system runs inside a protected, compliant setup.
Humberside Police — Humberside Police's Project Moriarty, a UK trial of AI in policing run with Coeus Software and the National Police Chiefs' Council (NPCC), projects Cecil could save over 23,000 officer hours annually per force. Frontline officers scored the police AI assistant's accuracy 4.6 out of 5; it gives verified procedural guidance in seconds.
Digital Jersey — A new body chaired by Digital Jersey, the Jersey AI Council, now steers AI adoption in Jersey. Its founders bridge public and private spheres: Government of Jersey, Jersey Financial Services Commission, Jersey Office of the Information Commissioner, and Jersey Finance, together promoting trustworthy AI uptake across the island.
Government of Hungary — Hungary's Resolution 1405/2025 mandates AI training for civil servants via a free course, 'MI Jövünk - Mesterséges Intelligencia alapjai', reaching staff at state-owned firms and state-funded universities alike. By 31 December 2026 the same decree demands AI teaching in schools and universities, plus vocational bodies. New bachelor's, master's and PhD tracks in artificial intelligence, tied to global and European research networks, are also required.
Government of Hungary — Hungary now requires an AI officer in every ministry under its AI governance in government ministries resolution, adopted 4 November 2025. Counties must name an AI-responsible official by 31 March 2026, and universities must assign AI specialists too, extending AI officer appointments in the public sector beyond central government. Civil servants and state-company staff face mandatory AI training, and by December 2026 ministers must fold AI teaching into education curricula.
Department for Work and Pensions — A newly built Department for Work and Pensions algorithm now scans two million Universal Credit journal messages every month, hunting for wording that suggests possible harm. Flagged cases turn red on staff screens, jumping the queue for quicker attention — this is the AI risk detection system for welfare messaging now running behind the scenes. Engineers trained this harm-detection model on 53,635 past urgent examples alongside 5.35 million ordinary ones, and it currently achieves a 79.7% recall rate.
Hungarian Parliament — Hungary now has Act LXXV of 2025, the national law implementing the European Union's Artificial Intelligence Regulation. This Hungarian AI regulation names the National Accreditation Authority as notifying body, hands market surveillance to the Minister for Enterprise Development, and sets up a Hungarian AI Council plus sector watchdogs, with the central bank covering financial AI systems. An AI regulatory sandbox and compliance procedures complete the EU AI Act implementation Hungary framework, mostly active from 1 December 2025.
Hrvatski sabor (Croatian Parliament) — Croatia has created a distinct AI system endangerment offence under Criminal Code Article 215a, covering danger to life, limb or property from an AI system, with six months to fifteen years by severity. Article 87(33) now defines 'AI system' per the EU AI Act, embedding AI-linked criminal liability in penal law.
Agenzia delle Entrate — Generative AI tax administration excludes machine drafting at Agenzia delle Entrate, Italy's revenue authority, barring AI-written assessments, refunds and litigation papers. Complaints over 'hallucinated' assessment notices sparked this internal prohibition, publicly confirmed 17 December. Approved systems still assist productivity or risk analysis under this AI-generated tax assessments rule, though staff must verify every result by hand.
National Assembly of the Republic of Slovenia (Državni zbor RS) — Slovenia's National Assembly adopted the Act on Implementation of the EU AI Act (ZIUDHPUI), the country's Slovenia AI regulation for artificial intelligence oversight. It names authorities to supervise AI systems, sets up a National Council for Ethics in AI, and requires regulatory sandboxes, building the nation's AI governance framework.
Državni zbor Republike Slovenije (National Assembly of the Republic of Slovenia) — Slovenia's National Assembly passed ZIUDHPUI, the Act on the Implementation of the EU Regulation on Harmonised Rules on Artificial Intelligence, Slovenia AI regulation naming enforcement bodies: AKOS, the Information Commissioner, the Bank of Slovenia, the Insurance Supervision Agency, the Market Inspectorate, and new notification bodies. It also creates a National Council for Ethics in AI with power over fines and penalties, part of Slovenia's harmonised AI rules oversight. Signed 31 October 2025, in the Official Gazette 6 November 2025, binding 21 November 2025.
National Agency for Fiscal Administration (ANAF) — Romania's ANAF now runs AI tax fraud detection spanning 11 linked databases, watching several risk signals together. Alexandru Nazare calls the AI-powered tax fraud software fully operational and 'state-of-the-art', noting peer agencies elsewhere in Europe already use similar tools. The rollout advances ANAF's digital overhaul, due to finish by the end of 2026.
State Customs Service of Ukraine — An AI customs bot now handles Ukraine's border traffic: the State Customs Service added a fallback tool for misread queries plus voice input so drivers and declarants can speak instead of type. Four AI-equipped mobile X-ray scanners also went live at the Poland border, widening automated border-control communication.
French government (DINUM) — DINUM's eight-month pilot gives 10,000 French civil servants in eight ministries an AI assistant for civil servants, built with Mistral AI on SecNumCloud-qualified infrastructure. It drafts emails, condenses documents and translates text. The trial tests a public-sector AI writing assistant against France's data sovereignty rules.
Digitaliseringsstyrelsen (Danish Agency for Digital Government) — Denmark AI regulation now has a six-part guidance set from the Danish Digital Government Agency, spelling out which artificial intelligence uses the EU AI Act bans outright. One document covers the full list of prohibitions; five companion texts drill into single cases: manipulation causing harm, exploiting vulnerable people, scoring citizens socially, reading emotions automatically. Each companion guide works as a variant entry point into the same banned AI methods.
Vonyarcvashegy municipality — Municipal chatbot Hungary has its first working example: Vonyarcvashegy switched on HALI, letting people handle routine administrative inquiries and get basic information from the town's website instead of visiting the local office. Tibor Navracsics unveiled this AI-powered assistant for local government, built for 32 languages and lighter workloads for customer service officers, calling it a turning point in state-citizen relations.
Danmarks Domstole (Danish Court Administration) — Every employee at Danmarks Domstole received Microsoft Copilot in courts access starting October 2025, though Danish policy on the tool lagged behind. A High Court judge criticised that gap, contending safeguards for this generative AI in courts belonged first, not afterward. The Court Administration confirmed the policy work continued.
Finnish Police — Finland's Police University College built the FERMI project's new AI disinformation detection tool for flagging fake news before it spreads. It maps how falsehoods travel online, weighs the danger each poses, and offers response options, letting commanders anticipate trouble and steer resources ahead of it. Pinpointing which patrols to deploy stays beyond its reach for now. The three-year effort united 17 European organisations, backed by EU Horizon 2021 funding.
Government of the Russian Federation — Russia's Ministry of Labour has enrolled over 450 federal officials in AI training for civil servants, part of a government AI skills programme rolled out nationwide. Instruction blends classroom sessions with remote-access tools, and successful participants earn a professional development certificate. Deputy Prime Minister Dmitry Grigorenko tied the effort to a simple aim: without trained staff, AI adoption across the public sector stalls.
Stadt Halberstadt — Startup fertig.ai built Patrick, an AI voice bot for citizen services, now fielding resident calls in Halberstadt. It handles multilingual questions on ID cards, passports and housing benefit applications, then routes callers to nearby providers like photo studios for passport pictures. Every exchange opens with disclosure that this is an AI-powered citizen inquiry assistant, and proceeds only after consent.
Landeshauptstadt Potsdam — Brandenburg's Potsdam now runs Wohni, an AI tool for housing benefit applications, after a positive trial. It verifies completeness and plausibility, trimming 21 percent off processing time and 20 minutes per case, with electronic file integration planned next. Officials call it the region's first such automated software in public administration.
Cabinet Office — One Big Thing 2025, the Cabinet Office's civil service AI training scheme "AI for All", gives every UK civil servant four 15-minute modules on artificial intelligence skills for government staff. It follows findings that just 28% felt confident using AI at work beforehand.
City of Přerov — Přerov deployed Ámos, an AI chatbot for citizen services, on its municipal website for resident queries in everyday language. nanotrix.ai, spun off the Technical University of Liberec, built it for 71,390 CZK yearly, pairing replies with source links. This virtual assistant now fields local government questions.
Government of Malta — Malta has named its Information and Data Protection Commissioner (IDPC) the market surveillance authority for high-risk AI systems under the EU AI Act, per Legal Notice 227 of 2025. This Malta AI Act designation gives the watchdog jurisdiction over deployments in policing, frontier control, immigration, courts, elections, biometric ID and emergency response — the categories the AI Act treats as high-risk. A companion clause in the notice lays out when live facial-recognition scanning of crowds may be authorised, forming part of Malta's AI regulatory compliance regime.
Parliament of Malta — Under Legal Notice 226, the Malta Digital Innovation Authority (MDIA) now leads enforcement of the EU AI Act domestically, a shift central to Malta AI regulation. Officials tasked the MDIA with market surveillance, notifications and acting as chief contact for AI Act queries, plus overseeing a sandbox for artificial intelligence trials. Breach penalties also fall within its remit, anchoring this Maltese AI governance framework.
State School for Public Administration (Republic of Croatia) — Croatia's State School for Public Administration launched AI literacy training for civil servants, teaching staff to speed routine work with AI tools while keeping people accountable for checking outputs. Institute for Development and International Relations specialists ran the session, "Fundamentals of AI Literacy: Practical Application of AI Tools." Croatia joins the earliest nations giving government staff a formal AI skills curriculum.
CSTAF (Conselho Superior dos Tribunais Administrativos e Fiscais) — Portugal's CSTAF is piloting AssessorIA, an AI tool for administrative court judges, in batches of 50-60 licences. It condenses case files, drafts sentence outlines, flags contradictions and picks out facts, an AI-assisted judicial decision support aid under a new ethics charter. Rulings still need a human judge's sign-off.
Swedish Government — A new law and ordinance for Sweden AI regulation has been put forward by the government inquiry AI-utredningen, aiming to fill gaps left by the EU AI Act. Its blueprint hands PTS the lead role in a national market-control system for overseeing AI, while IMY and Finansinspektionen would jointly supervise high-risk applications. Set out in report SOU 2025:101, titled "Anpassningar till AI-förordningen" (Adaptations to the AI Regulation), the plan further recommends a regulatory sandbox operated by PTS and stronger sanction powers for regulators, all due to take effect from 2 August 2026 under this Swedish AI law.
Urząd Miasta Gdyni (Gdynia City Hall) — Gdynia's city hall is Poland's first municipal government piloting PLLuM, the Polish Large Language Model, with the HIVE AI consortium for AI in citizen services. It will drive the local chatbot and bulletin search, helping residents find documents. Poland's Ministry of Digital Affairs expects similar AI-powered municipal services elsewhere.
Ministry of Industry and Trade of the Czech Republic — Czech AI regulation gains a draft bill: Czechia's Ministry of Industry and Trade wrote ten pages covering EU AI Act enforcement, pending inter-ministerial sign-off. Three watchdogs split duties — the Czech National Bank, the Office for Personal Data Protection and the Public Defender of Rights — plus the Czech Telecommunication Office as sole contact.
Government of Portugal — Anacom, the National Communications Authority, now serves as Portugal AI regulation's designated watchdog and contact for EU AI Act enforcement. Officials revealed the pick at the regulator's yearly gathering, citing its lengthy record overseeing telecom and tech markets. Lisbon finalised this weeks after Brussels' 2 August cutoff had lapsed.
Latvijas Investīciju un attīstības aģentūra (LIAA) — Latvia's Investment and Development Agency (LIAA) runs Maija, a generative AI virtual assistant that handles enquiries about its services and funding schemes non-stop, 24/7, via the LIAA site and business.gov.lv. The tool combines generative AI with a ready-made knowledge base, built atop Hugo.lv, the country's official language platform. This blended setup points to broader momentum behind AI-driven customer service across Latvia's public bodies.
Norwegian Digitalisation Agency (Digdir) — Norway's Digitalisation Agency (Digdir) has set up KI Norge, a national hub driving AI adoption public sector activity and private organisations across the Norwegian economy. The initiative pairs shared guidance with a regulatory sandbox run alongside Nkom and Datatilsynet, plus links to industry, universities and researchers, backed by NOK 30 million in government funding covering the hub, sandbox and supervisory function. Full operations for this artificial intelligence deployment in government begin in August 2026.
Italian Parliament — Italy AI regulation now includes a milestone: the Italian Parliament enacted the nation's inaugural comprehensive statute governing artificial intelligence, effective from 10 October 2025 and harmonised with the EU AI Act, positioning Italy ahead of every other EU member state on this front. Two bodies now oversee compliance, the Agency for Digital Italy (AgID) and the National Cybersecurity Agency (ACN), while lawmakers committed to refreshing a national AI framework every two years. Public funding worth one billion euros will back AI innovation alongside cybersecurity work and startup growth.
Ministry of Spatial Planning, Urbanism and State Property of Montenegro — Maša, an AI chatbot for construction permits and building legalisation, has gone live in Montenegro under the Ministry of Spatial Planning, Urbanism and State Property. Hosted on two fresh sites, legalizuj.me and gradjevinskadozvola.me, the tool walks residents through legalising unpermitted structures and lets investors apply for permits without visiting an office. Officials call it among the earliest rollouts of an AI assistant in public services across the Montenegrin government, with Minister Slaven Radunović pointing to lighter staff loads and faster real-time answers for the public.
Ministry of Taxes and Dues of Belarus (MNS) — Belarus's Ministry of Taxes and Dues put 300 employees, including 12 senior managers, onto Directum Ario for AI document processing. Across 2,675 tasks, 67% finished without human input, only 0.02% needed fixes, and workflow speed rose 45%. Novacom Systems built the tax-agency automation system.
Hessische Steuerverwaltung (Hessen Tax Administration) — Finanzamt Kassel produced Maxi, an AI use case for tax administration now operating across Hessen. Rather than officers checking statutes by hand, the system pairs case files with current legislation to draft compliant write-ups automatically. Trialled at Kassel from mid-September, this AI chatbot for tax officers expands nationwide, statewide, by year-end 2025.
Government of the Russian Federation — Artificial intelligence now flags problems in Russia's national projects before they occur, not after. Dmitry Grigorenko notes live indicator comparison, forecasts reaching 96% accuracy across over 1,000 scenarios per minute in this AI in government monitoring systems rollout. Officials, not the algorithm, retain final say in this AI-driven national project tracking effort.
Minister for Government Modernisation — Belgium's DeepSeek AI ban takes effect 1 December 2025 federally, ordered by the Minister for Government Modernisation. Belgian civil servants must delete DeepSeek applications and stop running them. Cybersecurity concerns drove the DeepSeek AI system prohibition: a June 2025 review by the Centre for Cybersecurity Belgium found troubling patterns in how the Chinese chatbot handles transmitted data.
Ministry of Labour and Social Affairs of the Czech Republic — Over 100,000 people monthly depend on Eva, Czechia's Azure OpenAI assistant for AI in government services, run by the Ministry of Labour and Social Affairs beside Jenda's AI Document Intelligence portal. Call centre load fell 60%, punctual payouts jumped from 60% to above 90% — a public-sector AI helpdesk win.
Ministry of Transport and Communications of Bosnia and Herzegovina — Bosnia and Herzegovina moves to fold AI adoption in government into its civil service, ditching paper-based procedures for verifiable digital ones. Edin Forto, Minister of Transport and Communications, plans workshops with UNDP on popular AI tools and alignment with European Union standards, part of a broader AI tools training for state officials push. Among 516 surveyed officials spanning 57 state bodies, UNDP found 69 per cent already lean on such tools daily, while 95 per cent had zero formal preparation beforehand.
Federal Public Service Finance (FOD Financiën) — Belgium's Federal Public Service Finance will apply AI tax return checks to Belgian company and individual filings, Finance Minister Jan Jambon said. Trials precede a wider launch, with AI-based tax compliance screening bound by privacy law and the EU AI Act. Figures and a start date are unspecified.
Government of Hungary — Government Resolution 1324/2025 sets Hungary AI policy, renewing the AI Strategy 2025-2030 and targeting 15% AI-driven GDP growth by 2030. The update aligns 2020's version with new EU AI Act requirements, adding yearly reviews. The Minister of National Economy and AI Government Commissioner lead the national AI roadmap.
Ministry of the Economy and Innovation (EIMIN) — €3 million in EU funding will back Lithuania's new AI regulatory sandbox, making it one of Europe's earliest adopters of such a scheme, per the Ministry of the Economy and Innovation. Small and medium enterprises gain a supervised space to build and refine AI tools before commercial launch, with the Innovation Agency and Communications Regulatory Authority (RRT) as watchdogs. The initiative reflects a wider Lithuanian AI policy push toward safe AI system development ahead of market entry.
Austrian Federal Administrative Court (BVwG) — AMS has won a GDPR fight over AI in employment services: Austria's Federal Administrative Court ruled AMAS doesn't breach Article 22's automation ban, since staff kept genuine power to overrule its verdicts. A 2020 Data Protection Authority ban on the automated decision-making system in public employment, in force since January 2021, is now void.
Gemeente Leiden — Gem now answers citizen questions on Leiden's municipal website 24 hours a day, an AI chatbot for municipal services drawing replies from the site's own pages. VNG Realisatie built this virtual assistant for citizen questions, named after 'gemeenten', and over twenty other Dutch municipalities already use it. Leiden is only trialling it, with permanent adoption depending on results.
National Agency for Information Society (AKSHI / NAIS) — Public feedback opened 28 August on Albania's five-pillar National Artificial Intelligence Strategy 2025-2030, built with World Bank backing. It aims to steer infrastructure spending, data rules, skills training and AI in the public sector while matching the EU AI Act, with rollout eyed for energy, procurement, tax-evasion checks, healthcare, farming and schooling, part of a wider public sector AI integration push. Formal approval is still pending.
State Employment Service of Ukraine — Ukraine's State Employment Service now runs an AI job search tool at dcz.gov.ua/aijob, filtering vacancies by city, salary or description with keyword lookups. Applicants submit CVs for AI-driven CV screening from a homegrown system suggesting openings, still in test mode. Resumes are wiped within seconds of matching, off outside servers.
Agencija za zaštitu osobnih podataka (AZOP) — Croatia's AZOP has published FAQs addressing AI data privacy compliance questions under GDPR, covering three practical areas: classifying when AI tools process personal data, assigning controller versus processor responsibility, and triggering mandatory impact assessments. This is applied artificial intelligence governance—existing law explained, not new legislation created.
Ministry of Economy of Ukraine — Ukraine's ePermit platform runs an AI veterinary licence verification module screening paperwork automatically, catching gaps before reviewers see them. Fewer applications get bounced, freeing specialists for harder cases. Built on Google DeepMind's Gemma model, funded by the EU, hosted on locked-down state servers instead of outside clouds — delivering AI-powered permit document checks nationwide.
Reykjavík City — Reykjavík City rolled out AI-powered public service search in August 2025, adding automatic summaries to its website's search function. Click-throughs to full result listings dropped 17.5%, since answers now display straight on-screen. An AI chat assistant, live since December 2024, logged almost 9,000 conversations across 2025, letting residents query services in Icelandic or English using natural-language AI search summaries.
Government of Luxembourg — Luxembourg's Bill No. 8476 sets out national authorities for EU AI Act implementation. The CNPD becomes the default market surveillance body, notified body for high-risk systems in law enforcement and immigration, and sandbox operator; ALIA takes on generative AI transparency oversight, forming the basis of AI Act compliance in Luxembourg. Submitted December 2024, the draft spent 2025 before the Media and Communications committee, drawing input from 12 bodies including the CNPD and Council of State.
Tiesu administrācija (Court Administration) — Latvia's Tiesu administrācija (Court Administration) built an AI anonymisation tool for court rulings that swaps names for tags like 'Person A', sparing clerks a tedious chore. Machine learning now scrubs identifying details from 411,005 civil, criminal and administrative judgments, easing what was once hand-done redaction across the courts.
Urząd Miasta Częstochowy (Częstochowa City Hall) — Częstochowa will test PLLuM, the Polish Large Language Model, under a pact between Krzysztof Gawkowski and Krzysztof Matyjaszczyk. Poland's Ministry of Digital Affairs picks the version and supplies the system, while officials use AI in local government to draft letters, condense paperwork and sift resident complaints. This marks generative AI adoption within Polish public-sector administration.
Ceļu satiksmes drošības direkcija (CSDD) — Latvia's Road Traffic Safety Directorate (CSDD) now runs an AI virtual assistant for vehicle registration, built by Dots since 6 August 2025 to speed up customer service replies. Staff skip manual searches through more than 2,000 regulatory documents and data files: the tool reads HTML, PDF, Word and image files, applying OCR to scanned paperwork. This reflects a wider shift toward AI-driven public service tools.
Parliament of the Czech Republic — Czechia's amended Act No. 110/2019 Sb. on Personal Data Processing, introduced through Act No. 230/2025 Sb., permits AI biometric identification police use at international airports for security screening in real time. Passed on 11 June 2025, the legislation labels qualifying technology an "isolated system", tying the definition to remote biometric recognition tools described under Article 3 point 42 of the EU AI Act. The measure narrows automated identity-scanning of travellers to airport security contexts rather than broader public surveillance.
Police of the Czech Republic — New judicial sign-off rules under the EU AI Act forced Czech officers to halt facial recognition airport surveillance at Prague's Václav Havel Airport. Running since 2019, the biometric setup had caught 155 people wanted on arrest warrants before shutdown.
Ministry of Justice — England and Wales will see AI in justice system tools launched under a new Justice AI Unit and Action Plan spanning courts, tribunals, prisons and probation, the Ministry of Justice confirms. Uses include staff transcription, prison capacity scheduling and public chat assistants. Ministers call the shift toward algorithmic courtroom support measured, not hasty.
Bundesagentur für Arbeit (Federal Employment Agency) — Germany's Federal Employment Agency now runs AI for company registration, an in-house language model that scans free-text business descriptions and suggests the five closest matches among 839 industry codes. Built by the agency's own statistics and IT staff, this industry classification AI took over from the previous manual search process. It launched 25 July 2025.
CNIL (French data protection authority) — France's CNIL, the French data protection authority, wrapped up guidance on AI regulation in France, explaining to developers how GDPR obligations bear on building models. Emerging from a GDPR compliance for AI systems consultation open to the public, the guidance tackles cutting down data gathered, narrowing its purpose, fixing retention periods, safeguarding individual rights, and grounding web scraping in legitimate interest. CNIL flagged that industry-specific rules would arrive afterward, in the second half of 2025.
Home Office — Britain's Home Office will pilot facial age estimation AI at border sites, scanning asylum seekers' faces to gauge whether they count as children or adults. Angela Eagle called the approach 'the most cost-effective option to pursue', with commissioning underway. The AI age verification method may enter routine checks by 2026.
Institute of Public Administration — Government AI adoption in Bulgaria scored 49 out of 100 in the country's first AI Readiness Index, published by the Institute of Public Administration. The public sector AI readiness rating, drawn from polling 6,254 civil servants in April and May 2025, places the administration in a 'building capacity' phase. Openness and learning motivation stood out as strengths, while strategy, leadership and AI Act preparedness lagged.
Ministry of Digital Governance — The Ministry of Digital Governance houses a Special Secretariat for Artificial Intelligence and Data Governance (Joint Ministerial Decision 1961/2025, Government Gazette B' 3804). Its job: shaping, coordinating and delivering Greece AI policy plus public data governance. Via Directorates for AI and Data Governance, it voices national AI policymaking on the EU AI Board and in international AI fora.
Najvyšší kontrolný úrad SR (NKÚ) — AI audit automation now compiles final audit reports at Slovakia's NKÚ from individual audit protocols. Its agenda platform also runs a secure chatbot, document summarisation and meeting notes generation, on the Interior Ministry's government cloud with OpenAI services via MIRRI — an AI-driven workflow now standard, not trialled.
Data Protection Commissioner (Komisioneri për të Drejtën e Informimit dhe Mbrojtjen e të Dhënave Personale) — Guidance No. 05/2025 on the Processing of Personal Data by Competent Authorities turns Law No. 124/2024 into working practice across Albania's police, prosecution and judicial bodies, zeroing in on AI-driven facial recognition law enforcement activity. The Data Protection Commissioner now insists on impact assessments, logged activity trails and internal whistleblowing channels wherever behavioural profiling or biometric identifiers come into play. Justice-system agencies also face fresh curbs on automated decision-making, widening scrutiny of biometric data processing by police.
Belgian Federal Government — Belgium's federal government now runs on a shared ethics code: forty public organisations back the new Charter for Responsible AI Use in Government Services, a project Minister Vanessa Matz put forward. Fairness, transparency and human oversight anchor the rules guiding artificial intelligence rollout across federal government AI. The framework aims to steer how public sector artificial intelligence gets used nationwide.
Hessisches Polizeipräsidium (Hessen State Police) — Germany's Hessen state now leads the country in AI video analysis in policing, running the technology on live feeds across Frankfurt's Bahnhofsviertel district after lawmakers revised state police legislation. Warrants issued by judges permit officers to sift footage for missing people, abduction victims and suspected terrorists.
Czech government — Chinese law forces DeepSeek to cooperate with state authorities there, exposing user data to unauthorised access, so Czechia's National Cyber and Information Security Agency (NÚKIB) pushed for an outright block. That advice became a nationwide directive restricting AI use in government agencies, shutting the Chinese firm's tools out of every ministry, an AI ban in public sector systems targeting one named vendor rather than AI broadly.
Bank of Russia — Russian financial institutions now have Bank of Russia guidance on AI in finance: a Code of Ethics built on five principles — human-centricity; fairness; transparency; security; and, separately, responsible risk management. Firms should disclose AI use to clients, allow opt-outs, audit quality, protect data and manage risk. These AI systems in the financial market guidelines remain non-binding.
Føroya Landsstýri (Government of the Faroe Islands) — Ruth Vang's Ministry of Finance now owns AI policy area responsibility in the Faroe Islands, folding it into the existing IT brief. Regulations confirm the shift, placing artificial intelligence governance among the ministry's standing duties rather than treating it as a passing research topic.
San Marino6 Jul 2025Guardia di Rocca — San Marino's Guardia di Rocca has rolled its own: Atena, a homemade AI chatbot for criminal investigations. Officers lean on OSINT (Open Source Intelligence) tradecraft, letting the assistant trawl public sources and lighten heavy caseloads tied to cybercrime. Since roughly April 2025 this AI-powered OSINT investigations build has run free of added expense.
Vláda Slovenskej republiky — On 4 July 2025 Slovakia named Radoslav Štefánek as its first Government Plenipotentiary for Artificial Intelligence, steering Slovakia AI policy across ministries, regional bodies, companies and universities while representing the country in international AI forums. A cabinet resolution created the post, run as an advisory arm inside MIRRI, where Štefánek already holds the State Secretary title. This reflects a national AI coordination model rather than a standalone ministry.
Bundesnetzagentur (BNetzA) — Germany's Federal Network Agency (Bundesnetzagentur) now runs a KI-Service Desk helping firms and public bodies through EU AI Act compliance. Its Compliance Compass tool flags whether an AI system sits within regulatory scope, faces transparency rules, or ranks as high-risk, part of wider AI regulation guidance nationally. Free AI literacy courses, required under Article 4 since 2 February 2025, complete the service.
Government of Iceland (Ministry of Culture, Innovation and Higher Education) — 20 fully financed measures anchor Iceland's Action Plan, an Iceland AI policy grouped into five priority areas: societal benefit, business competitiveness, education, reimagined public services, and healthcare. Rolled out in 2025 as a national AI strategy, it spans two years, undergoes six-monthly checks, and faces full assessment by mid-2027. Logi Einarsson framed it as vital for a small, technologically advanced nation pursuing its potential.
Czech Ministry of the Interior — Czechia's Interior Ministry is piloting an AI assistant for administrative hearings at chosen offices to shorten processing time. Staff will use the AI tool for administrative proceedings to review files, legislation and precedent, then generate paperwork automatically. Financed with 50 million CZK from the National Recovery Plan, the project runs to August 2026.
HM Customs Gibraltar — AI risk assessment customs controls now govern Gibraltar's Commercial Gate cargo, built into ASYCUDA Version 4.4. Four tiers of intelligent selectivity - automatic release, documentary checks, physical inspection, post-clearance audit - ease compliant traders through while directing scrutiny to riskier consignments. HM Customs calls this AI-driven customs clearance faster for clean compliance records.
Government of Norway — Norway's Ministry of Digitalisation and Public Governance launched a three-month public consultation on Norway AI regulation, transposing the EU AI Act into a draft domestic statute. Officials hope KI-loven, the Norwegian AI Act, takes effect by summer 2026, aligning with the EU AI Regulation's main rollout.
Czech government and Microsoft — Czechia aims to give more than 350,000 residents artificial intelligence training within a year through the AI National Skilling Plan, a joint venture between the Czech government and Microsoft. Backing comes from Microsoft, which is contributing over CZK 10 million in first-phase funds, reaching civil servants, teachers, students, judges and jobseekers under this AI training in government drive. Linked to the National AI Strategy 2030, the scheme advances the nation's broader AI skills development goals.
Landsverk (Faroese Public Roads Administration) — Since June 2025, Landsverk, the Faroese Public Roads Administration, has run an AI road damage detection system flagging holes, cracks and surface faults on national roads. Staff rank repair urgency instead of inspecting manually. Analyst Hilmar Ellingsgaard Petersen said this speeds road-surface deterioration identification.
Council of Ministers of Bulgaria — EU AI Act Bulgaria implementation now assigns fundamental rights protection to seven named authorities, following Decision No 398, adopted by the Council of Ministers. Oversight duties spread across the Ombudsman, the Central Election Commission, the Commission for Protection from Discrimination, the Commission for Personal Data Protection, the Commission for Consumer Protection, the State Agency for Child Protection, and the General Labour Inspectorate Executive Agency. Each watchdog now enforces AI Act compliance within its existing legal remit, marking a concrete step toward meeting these obligations nationally.
Dátueftirlitið (Faroese Data Protection Authority) — Handling personal data through ChatGPT breaches the Data Protection Act, Dátueftirlitið, the Faroese Data Protection Authority, cautioned, since nobody can confirm how such tools retain or secure information. This is core to sound AI in public-sector case processing, it said, urging public bodies to set their own rules before adopting generative AI in government data processing.
NAV (Norwegian Labour and Welfare Administration) — Every advisor at Norway's NAV Kontaktsenter (NKS) can now call on Bob, an AI assistant covering this AI customer service function. The RAG-based system mines roughly 1,100 welfare benefit articles so staff answer citizen questions quicker, serving as an AI contact centre assistant. Uptake climbed from a 30-advisor trial to 120 before full rollout.
Isle of Man Information Commissioner — Six years had passed since the last review from the Isle of Man Information Commissioner's Office when its 2025 report named artificial intelligence and high-risk personal data handling as a growing focus under AI data protection regulation. Government bodies alone accounted for all four Data Protection Impact Assessments submitted that year, with not a single filing from private firms, a pattern the regulator took as evidence that businesses remain unclear on their legal responsibilities toward automated systems. Commissioner Alexandra Delaney-Bhattacharya pressed organisations to grasp that any AI tool processing personal data is subject to Manx law, reinforcing calls for stronger AI regulation compliance island-wide.
Government of the Russian Federation — A new Centre for AI Development now sits under Russia's federal government, formed by Resolution No. 861 atop the existing Analytical Centre — the latest move in Russia AI regulation. Deputy Prime Minister Dmitry Grigorenko leads it, with duties covering AI governance in Russia: picking and scaling standard tools, linking ministries, regions and firms, giving methodological guidance, and handling international cooperation.
Prokuratura Okręgowa Warszawa-Praga — The Warsaw-Praga District Prosecutor's Office confirmed on 4 June 2025 it is using AI video evidence analysis on an investigation into the far-right group Rodacy Kamraci, covering over 1,700 recordings totalling more than 3,000 hours. Officials called the software modern AI-based tools for reviewing video footage.
Estonian National Audit Office — Estonia's National Audit Office found only about 30 AI-based solutions actually running in the public sector, far short of the 130 systems logged in the Ministry's Kratt register — a stark gap in AI adoption in government. Most tools remain under trial, not yet serving citizens. Weak awareness, shaky data and legal hurdles were named as chief brakes on wider AI in the public sector.
Government of the Czech Republic — Under Resolution No. 393, Czechia runs AI governance Czechia via a Government Commissioner for Artificial Intelligence, linking ministries, public bodies and the AI ecosystem. Duties span the NAIS Action Plan, matching rules to the AI Act, and building bodies for artificial intelligence oversight. Jan Kavalírek was first appointed; Lukáš Kačena took over on 12 January 2026.
RTR GmbH / Austrian Federal Government — RTR's KI-Servicestelle built an AI Act chatbot answering public queries on the EU AI Act in German, giving Austria a public AI regulation assistant. It uses retrieval-augmented generation (RAG) to draw answers only from verified sources, including the Act's text. The goal: making the compliance framework understandable for citizens, organisations and companies.
Government of Luxembourg — A Luxembourg AI policy sits inside "Accelerating Digital Sovereignty 2030", alongside linked data and quantum plans. Funding for 2025-2030 targets hands-on trials and real-world use across industries. The Luxembourg AI strategy also promises swift, pragmatic EU AI Act rollout via regulatory sandboxes.
Ministry of Digital Transformation of Ukraine — A new scheme gives Ukraine's public workforce, from senior officials to IT staff, access to AI training for civil servants, backed jointly by the Ministry of Digital Transformation, Google and CDTO Campus. Under the banner Artificial Intelligence for the Public Sector, the free online course builds skills in responsible AI adoption in the public sector, automating repetitive work, and writing sharper prompts. The Ministry of Economy was the pilot agency, kicking off its cohort in June 2025.
Finanzverwaltung Nordrhein-Westfalen (NRW Finance Administration) — Eight tax offices across North Rhine-Westphalia now handle AI tax assessment, doubling the four where the finance administration first switched it on. Its home-built module scans filing patterns, sorting straightforward, low-risk returns for hands-off processing so staff can focus on trickier files. Running on the KONSENS federal framework, this AI income tax processing effort is set to reach the entire state during 2026.
Folketinget (Danish Parliament) — Denmark AI regulation now runs through Law No. 467, the Act on Supplementary Provisions to the Regulation on Artificial Intelligence, passed by the Folketing and active from 2 August 2025. Three bodies enforce it: Digitaliseringsstyrelsen (Danish Agency for Digital Government) as lead contact, Datatilsynet (Danish Data Protection Agency) and Domstolsstyrelsen (Danish Court Administration), which inspect firms, demand records and levy fines (bødeforelæg) for breaching the EU AI Act — Denmark's version of EU AI Act enforcement.
Folketinget (Danish Parliament) — Denmark AI regulation now sits with three bodies holding fresh powers to probe firms and levy fines. Law no. 467 hands the Danish Agency for Digital Government, the Danish Data Protection Authority and the Danish Court Administration authority over the EU Artificial Intelligence Act, setting AI Act compliance duties from 2 August 2025.
Ministry of Finance of the Czech Republic — Adam, a chatbot built by CIIRC CTU, now handles AI in subsidy auditing for Ministry of Finance staff checking grant claims. It answers plain-language queries and cites sources, gaining access to non-public files from June 2025 as this grant compliance assistant's reach grows.
Cabinet of Ministers of Ukraine — Ukraine's Cabinet of Ministers adopted the Action Plan for Implementing the Concept for the Development of Artificial Intelligence in Ukraine for 2025-2026, a binding roadmap for Ukraine AI regulation. It covers drafting an AI law, cybersecurity guidance, higher-education standards, procurement analysis tools, defence uses, and aligning domestic rules with international AI policy in Ukraine frameworks. The plan is mandatory for relevant ministries and agencies.
Government of Finland — Government Proposal 46/2025 now sits before Finland's Parliament, establishing an Act on the Supervision of Certain AI Systems that transposes the EU AI Act into domestic law — the core of this Finland AI regulation. Rather than one central regulator, several market surveillance bodies split enforcement duties, with the Finnish Transport and Communications Agency, Traficom, designated as sole point of contact for Brussels. A dedicated Sanctions Board handles penalties topping €100,000, leaving lesser fines to separate authorities under the wider AI Act compliance framework.
Department of Public Expenditure — Ireland's Department of Public Expenditure launched a framework covering AI in public service: the "Guidelines for the Responsible Use of Artificial Intelligence in the Public Service". Jack Chambers presented it with a Decision Framework, a Responsible AI Canvas Tool, and free staff training. This reflects Ireland's push for responsible AI adoption in government, favouring guidance over mandates.
Financial Administration of the Republic of Slovenia (Finančna uprava RS) — Slovenia's FURS built an AI chatbot for tax information. It covers VAT and fiscal cash registers, reachable 24/7 as a generative AI tax assistant.
Federal Data Protection and Information Commissioner (FDPIC) — Existing law already covers AI in Switzerland, the Federal Data Protection and Information Commissioner (FDPIC) has ruled: the Federal Act on Data Protection (FADP) applies directly to AI-supported data processing, no new statute required. Under this Swiss AI regulation, anyone building, offering or deploying an AI system must explain what it does, why, and where its data comes from, and people keep the right to know they're talking to a machine, not a person. High-risk processing can be justified with adequate safeguards, but the same data protection AI compliance regime rules out real-time facial recognition and social scoring outright. FADP has been in force since September 2023.
Krajowa Szkoła Administracji Publicznej (KSAP) — KSAP trained Polish civil servants in AI in public administration through a 2025 series: prompt engineering, drafting with ChatGPT and Microsoft Copilot, ethical artificial intelligence use in government. "AI in the Office" opened the run as one online day, plus sessions on 30 June, 8 October and 13 October.
Malta Tax and Customs Administration — Software now flags likely evaders for Malta Tax and Customs Administration, recovering €400 million in unpaid tax through AI tax evasion detection. Clyde Caruana linked the sum to a €650 million year-on-year jump in income tax, catching evasion before gaps could be used. Automated compliance tools reshape revenue pursuit.
Løgtingið (Faroese Parliament) — The Faroese Parliament enacted Law No. 78, anchoring AI in public sector services within Digital Faroe Islands 2.0, a 2025-2027 blueprint. Building atop the earlier Talgildu Føroyar scheme, the Ministry of Finance's rules task the Vangin portal with answering residents' private questions via AI-driven public administration tools drawing on official records.
Government of Croatia (Ministry of Justice, Public Administration and Digital Transformation) — The government is drafting a National Plan for the Development of Artificial Intelligence to 2032, plus a 2026 to 2028 Action Plan. Three pillars frame this Croatia AI policy: public policy, investment, conditions for effective AI adoption—education, economic digitalisation, research and innovation support, citizen data protection. Experts drafting the national AI policy first met on 27 May 2025.
Commission for Personal Data Protection (CPDP) — Bulgaria's CPDP has barred shops from using facial recognition AI in retail stores to identify customers, ruling that no GDPR exemption, including theft prevention, permits processing biometric data this way. Because the technology relies on AI, the AI Act also applies, classifying such remote biometric identification technology in shops as high-risk.
Data Protection Commissioner (Komisioneri për të Drejtën e Informimit dhe Mbrojtjen e të Dhënave Personale) — The Data Protection Commissioner's Instruction No. 3 on the Processing of Personal Data through Video Surveillance Systems binds state and commercial CCTV operators, placing facial recognition and video analytics under Albania AI regulation. These AI surveillance rules keep cameras out of bathrooms and changing rooms, forbid live-streaming identifiable footage, delete recordings after 72 hours (extendable to 30 days), and demand notices plus data protection impact assessments.
Ålands landskapsregering — The Government of Åland has adopted an AI policy for public administration, setting rules on permitted AI uses in its agencies. Deployments need a prior risk assessment, and invoice processing is named as a priority use case for pattern recognition cutting manual work. This is Åland's first formal AI governance policy for government operations.
Cabinet Office / DSIT — Redbox, an AI assistant for civil servants from the Incubator for AI, serves over 2,000 UK staff at the Cabinet Office, No 10 and the Department for Science, Innovation and Technology. It condenses papers, drafts briefings and polishes text using GPT-4, Claude and Gemini. Cleared to OFFICIAL SENSITIVE, it signals a wider civil service AI rollout ahead.
Gemeente Goes — Guus 3.0, Goes's third-generation AI chatbot for citizens, restricts replies to approved municipal sources. It runs on servers in the Netherlands and Sweden, purging data automatically and excluding it from training. This citizen-facing virtual assistant draws only on vetted content, not the open web.
Dutch government (Ministry of the Interior and Kingdom Relations) — Dutch civil servants can now use generative AI in public sector work more freely under new government-wide rules covering municipalities, provinces, water authorities and executive bodies. The policy requires risk analyses beforehand and favours European or open-source tools over other options. This shift toward broader public sector AI adoption gives officials clearer conditions for deploying such tools in daily work.
Agència Andorrana de Protecció de Dades — A formal probe by the Agència Andorrana de Protecció de Dades, Andorra's Data Protection Agency (APDA), targets Meritxell, the government's AI assistant answering citizen questions on the EU association agreement, examining whether personal data was properly safeguarded in this government AI assistant case. Officials, through spokesperson Guillem Casal, denied storing any personal information and called the scrutiny unexpected, though they admitted an earlier warning had flagged a privacy-policy gap on the official portal, now resolved. At stake: AI data compliance standards for public-sector chatbots.
Oxfordshire County Council — 81% of Oxfordshire County Council's staff now use Microsoft Copilot, a milestone in AI adoption in local government. Workers reclaim 27.7 minutes daily via the assistant, applying it to demographic analysis, social content drafting and event organisation — a hands-on AI tool for council staff.
Bundesstrafgericht (Federal Criminal Court) — Switzerland's Federal Criminal Court in Bellinzona trials AI court transcription, using software to draft automated hearing logs and mask names in rulings automatically. The pilot also tests whether the tool can speed up reviewing seized devices, scanning PDFs on phones and laptops gathered as evidence.
Isle of Man Constabulary — AI image grading lets Isle of Man Constabulary classify indecent images of children by matching them against verified entries in the UK's Child Abuse Image Database (CAID), following the territory's adoption of England and Wales sentencing categories from 21 April. Jane Poole-Wilson said the automated classification of abuse imagery limits officers' exposure to distressing material and speeds up cases.
Government of the Czech Republic — Czechia's Czech AI policy now channels CZK 19 billion through an Action Plan tied to the National AI Strategy 2030, run by the Ministry of Industry and Trade. Seven areas share the money: research, education, industry, public administration. A national artificial intelligence roadmap this size marks the country's biggest implementation step yet.
Comunidad Autónoma de Galicia — Galicia's Ley 2/2025 stands as the regional AI law Spain now counts as its first, a single rulebook steering how artificial intelligence gets built and used by government bodies and businesses alike within the community. Under this AI governance framework Galicia now requires civil servants and ordinary residents to receive training in understanding AI systems. The statute folds an older blueprint, the Estrategia gallega de inteligencia artificial 2030, into enforceable legislation.
President of the Republic of Belarus — Presidential Decree No 135 ranks 'digital technologies and artificial intelligence' first of five national priority directions for scientific, scientific-technical and innovative activity, setting Belarus AI policy for 2026-2030. Every state scientific programme and the State Innovation Development Programme must steer public R&D funding toward AI priorities and kindred digital technologies.
Romanian Parliament (Senate) — Romania's Senate has registered legislation setting rules for AI regulation in sensitive fields such as healthcare, justice, and public administration, framed around the responsible use of artificial intelligence. Courts would be barred from letting algorithms decide cases, doctors must sign off on any AI-generated diagnosis, and consent becomes mandatory before AI can build political profiles; manipulative deepfakes aimed at swaying public opinion are banned outright. Public bodies deploying such systems face transparency duties and periodic audits, while breaches touching fundamental rights risk fines reaching 100 million lei (20.1 million euros) plus two to five years' imprisonment.
Federal Ministry of the Interior and Community (BMI) — Germany's BMI now issues nationwide guardrails covering AI in federal administration, via its 'Leitlinien für den Einsatz Künstlicher Intelligenz in der Bundesverwaltung'. Agencies gain shared principles for rolling out such systems safely and consistently — part of broader traction for AI governance in the public sector.
Finnish Immigration Service (Migri) — Finland's Migri auto-approves extended researcher permits once conditions are met: a master's degree, a Finnish research organisation post, and a complete online submission. Only positive outcomes are machine-generated under this AI in immigration permit processing system; harder cases still go to a specialist. Migri calls this an automated visa decision system.
Government of Norway — Nkom (Norwegian Communications Authority) leads Norway AI regulation, appointed as coordinating supervisory body over EU AI Act compliance for Norway's Ministry of Digitalisation under Karianne Tung. Firms, consumers and agencies gain guidance from Nkom, while Datatilsynet and Digdir join a shared sandbox advancing AI oversight nationwide.
Swiss Federal Council — Three pillars anchor Switzerland AI policy under the KI-Teilstrategie, the Federal Chancellery's Strategy for the Use of AI Systems in the Federal Administration, briefed to the Swiss Federal Council: sharpening staff skills, winning public confidence through lawful, ethical practice, and lifting output by automating repetitive tasks. A subsequent plan layers on eight concrete measures with a 2026-2028 roadmap, extending AI regulation uniformly across federal agencies.
Swedish Government — A new law would let Sweden's police run AI facial recognition police checks in real time against people in public spaces, part of wider biometric remote identification powers under memorandum Ds 2025:7. Officers would normally need a court's or prosecutor's sign-off first, except in urgent cases, where approval can follow within 24 hours. Swedish consultation on the plan produced a formal government proposition in March 2026.
Governorate of Vatican City State — Vatican City State has stood up a five-member Commission on Artificial Intelligence for AI governance, created under Decree N. DCCII containing the Guidelines on Artificial Intelligence by Sister Raffaella Petrini. Seats went to the Legal Office, the Directorate of Telecommunications and Information Systems, and the Directorate of Security and Civil Protection Services. Ahead: drafting AI regulation, weighing AI deployments, warning of hazards, and issuing twice-yearly findings on AI's reach.
Gendarmerie Nationale and ANFSI — Parole, ANFSI's AI transcription in law enforcement tool at the French Gendarmerie, a speech-to-text system, cuts interview time from eight hours to two-four hours via retrained OpenAI Whisper. Family protection units use it for minor victim sessions. Other Gendarmerie AI: sorting child sexual abuse imagery, I-Accueil for public queries, PredNatinf for penal code lookups.
Kanton Solothurn Steueramt — Solothurn's tax office now runs AI tax assessment across all 175'000 natural persons' tax returns filed in the canton, using automated tax return processing that sorts cases into those it can close alone and those sent for human review. It follows Obwalden and Nidwalden as the third Swiss canton to automate this way, targeting a fifth of assessments finished with zero staff input. Thomas Fischer said the goal is easing hiring shortages, not cutting headcount.
Government of Spain — Spain's government greenlit an AI governance bill aligning domestic policy with the EU AI Act, promoting ethical, inclusive and beneficial use of artificial intelligence. New digital rights let authorities pull AI systems from the Spanish market after a serious incident, backed by fines from €7.5M to €35M for banned practices. This marks an early step toward a broader AI regulation framework.
Copenhagen Police (Københavns Politi) — Facial recognition AI in policing gets a real trial: Copenhagen Police and the Danish National Police now scan video evidence from criminal probes with this video evidence analysis technology. Object recognition covers all case types; facial matching stays limited to serious crimes like murder and rape. The pilot runs to autumn 2025, aiming to speed up review of large footage volumes.
Saeima (Parliament of Latvia) — A new legal basis for AI regulation in Latvia arrives via the Artificial Intelligence Centre Law, passed by the Saeima, which also stands up a national AI Centre due to open by 31 March 2025. Tasked with linking government, industry and universities, the body will foster AI skills, manage risk and may operate a sandbox for trialling AI systems under real conditions. Its founders — VARAM, the Ministry of Economy and the Ministry of Defence — bring in private companies and academic institutions as partners.
Czech Ministry of Justice — Czechia's Ministry of Justice now runs Microsoft 365 Copilot Chat internally, giving officials a tool for daily paperwork, data review, slide-building and replies to citizens, part of wider AI adoption in government administration. Because the system stays closed off from the open internet, citizen records remain inside a secure EU Data Boundary rather than leaving it. Justína, an earlier assistant from the same ministry, had already reached more than 57 058 users in five months, a public sector AI rollout that preceded this one.
Government of Andorra — Andorra piloted AI for public administration in 2025, testing a Ministry of Finance tool for tax checks alongside a natural-language search engine for citizen queries. Marc Rossell said results fell short across both trials, pointing to weak data quality as the root problem, and neither moved beyond internal testing into AI in government services for public use.
Government of Ireland — Eight bodies will share EU AI Act implementation in Ireland instead of one super-regulator: the Central Bank of Ireland, the Data Protection Commission and the Health Products Regulatory Authority among them. Each already polices AI regulation in its own sector. A new AI Office of Ireland coordinates centrally from August 2026.
National Police Chiefs' Council — AI policing gets a national roadmap in England and Wales: the National Police Chiefs' Council's AI Strategy 2025 seeks global leadership in responsible artificial intelligence for stopping, spotting and probing crime. Governance, ethics and milestones to 2027, including a national AI lab, support the goal.
States of Jersey Police — Scott Hall confirmed a pilot of AI transcription for witness interviews at States of Jersey Police, citing UK ethical guidelines. He called early results "significant dividends" for the force, part of a wider move toward automated speech-to-text tools in policing.
National Academy of Sciences of Belarus — Belarus's National Academy of Sciences (NAS) has opened the Scientific Artificial Intelligence Center (AI-center), an artificial intelligence research centre built from the United Institute of Informatics Problems' digitalisation office, with a partner firm from the China-Belarus Industrial Park 'Great Stone'. Now the country's main AI research and development hub, it covers cybersecurity, supercomputing, neural network coding, image recognition and language-processing systems, serving as Belarus's core base for both theoretical and applied AI work.
Ministru kabinets (Cabinet of Ministers) — Latvia AI regulation now hands lead authority to the Ministry of Smart Administration and Regional Development (VARAM) for the EU AI Act. Three bodies share market oversight — Consumer Rights Protection Centre, State Data Inspectorate and Bank of Latvia — while the Ombudsman covers fundamental rights. Cabinet approved this EU AI Act implementation plan.
Government of Hungary — Resolution 1028/2025 gives Dr László Palkovics Hungary's AI policy brief as mesterséges intelligenciáért felelős kormánybiztos (government commissioner for artificial intelligence). Remit: defence, health, trade, education, energy, transport, public administration, science, law enforcement, agriculture, financial sector, economy, industry and national security. AI governance duties include founding and chairing the Hungarian AI Council.
APDP (Autorité de Protection des Données Personnelles) — Monaco's Personal Data Protection Authority joins 61 other regulators in a joint statement on AI image generation tools producing lifelike, identifiable images and videos without consent. Signatories demand stronger safeguards, real transparency, working takedown routes, and child protections against this AI-generated deepfake harm.
Swiss Federal Council — Ratifying the Council of Europe's AI Convention has won out over a unified AI regulation in Switzerland, the Federal Council confirmed. Field-specific rules will apply to healthcare and transport, while wider protections stay confined to fundamental-rights topics like data protection. Departments must deliver a consultation draft by the end of 2026, so Swiss artificial intelligence law seems improbable before 2028.
Imprensa Nacional-Casa da Moeda (INCM) — Portugal's INCM built Lia, a generative AI chatbot for government gazette Diário da República. Residents can now pose everyday questions about consolidated statutes and receive instant replies, covering eligibility for remote-work schemes or housing aid. Generative AI for legal information access now reaches the general public directly.
Ministry of Digital Affairs — Poland's Ministry of Digital Affairs has reworked its draft Act on Artificial Intelligence Systems after weighing over two thousand consultation comments. This Polish AI regulation creates the Commission for AI Development and Safety (KRiBSI) as watchdog, lets micro, small and medium enterprises join sandboxes free, and adds individual opinions under the AI Systems Act, giving firms certainty pre-deployment.
National Data Protection Commission (CNPD) — AI regulation Luxembourg materials from CNPD brand most unacceptable-risk systems under the EU Artificial Intelligence Act unlawful, arguing these prohibited AI systems generally violate GDPR too. Flagged practices span social scoring, predictive policing, emotion tracking on the job and harvesting facial images without consent, leaving Luxembourg-based deployers exposed to inquiry.
Gouvernement français (Comité interministériel de l'intelligence artificielle) — Under France AI policy, Prime Minister François Bayrou has unveiled a third stage of national AI strategy titled 'Faire de la France une puissance de l'IA'. Agreed at an interministerial gathering, the blueprint sets four aims: expanding computing and data-centre capacity, drawing skilled workers via dedicated visas and research chairs, embedding generative AI tools inside every government ministry, and spreading adoption across the wider French economy through the Osez l'IA scheme. This follows two earlier rounds of the same push to make the nation a leading force in artificial intelligence.
Personal Data Protection Office (UODO) — The Polish Personal Data Protection Office (UODO) advises extreme caution before anyone uses DeepSeek R1, a Chinese AI chatbot. UODO's AI chatbot data protection concern centres on storage: information may sit in China, within reach of Chinese authorities. The regulator is assessing, on chatbot privacy grounds, whether DeepSeek's data processing meets EU requirements.
France Travail — AI in employment services now underpins two adviser-facing systems that France Travail, the national employment agency, has launched alongside Mistral AI. Advisers can draw on ChatFT to help draft written material and on MatchFT to pair jobseekers with suitable vacancies, extending a setup that already runs scheduling and coaching support through generative AI matching tools for close to 10,000 agents and users.
Tribunal de Contas (Court of Auditors) — AI in public procurement auditing has moved from plan to practice at Portugal's Tribunal de Contas (Court of Auditors), whose new model took two years to build with the European Commission and OECD. Precise indicators now single out which public contracts merit closer scrutiny, automating the first screening pass so staff spend their attention on the riskiest files. It stands as a live rollout of AI-driven risk assessment for public contracts, meant to strengthen oversight of government spending.
Belgian Federal Government — The Belgian Institute for Postal Services and Telecommunications now holds the AI Act market surveillance authority role, fixed by the Government Declaration and reaffirmed in the 2025-2029 Federal Government Agreement. This AI Act enforcement authority spreads fundamental-rights supervision across twenty-one bodies under Article 77, the highest count among EU member states.
Valsts ieņēmumu dienests (VID) — VID, Latvia's State Revenue Service, piloted Vidis, a generative AI assistant for tax consultants, built by Latvian firm Tilde on retrieval-augmented generation (RAG). It pulls guidance from legal texts and FAQs so consultants can answer taxpayer questions faster. Over half the outputs passed as correct in round two of testing — an early signal for AI in tax administration, not proof of a finished, audited system.
Flemish Government — Flanders will hand 10,000 Flemish civil servants and local staff Microsoft Copilot licences under Europe's biggest public sector Copilot deal. Prime Minister Matthias Diependaele sealed the four-year pact at Davos, betting on quicker paperwork and admin tasks via AI-assisted government services, part of Belgium's push toward wider public-sector AI adoption.
Seimas of the Republic of Lithuania — Lithuania AI regulation gained a legal footing when the Seimas passed Act XV-105, aligning the Law on Technology and Innovation with Regulation (EU) 2024/1689. The Innovation Agency now serves as notifying authority for AI conformity assessment, a statutory sandbox for testing AI systems was established, and start-up classification thresholds were narrowed. Most articles applied from 1 April 2025.
Government of the Republic of Serbia — Adopted by Serbia's government, the Strategy for Development of Artificial Intelligence in Serbia 2025-2030 replaces the country's 2020-2025 predecessor and drives Serbian AI policy toward stronger legislation, education funding, innovation, infrastructure and broader public-sector use of AI. Wider AI deployment across government agencies sits alongside efforts to match global technology trends and support the 'Leap into the Future 2027' programme.
Ministry of Justice, Public Administration and Digital Transformation — ANON, Croatia's new AI court decision anonymisation system, strips identifying details from rulings across every court and posts them to one searchable public portal. Linked to eSpis, it pulls each ruling 15 days after dispatch for automatic redaction and metadata indexing, retiring manual redaction of judicial rulings nationwide.
EU AI Act bans in force since 2 February 2025, and unaffected by the 2026 delay to high-risk rules, cover predictive policing based only on profiling or personality traits and facial recognition databases built by untargeted internet or CCTV scraping. Police use of live remote biometric identification in public places is also barred outside narrowly defined purposes.
Police live facial recognition in London led to more than 1,300 arrests between January 2024 and September 2025. The National Physical Laboratory found no statistically significant performance gaps by age, gender or ethnicity at UK forces' settings. The Home Office publishes this on its own policy; it holds only with correct settings.
Council of Ministers of the Republic of Belarus — Belarus's Council of Ministers approved the Concept of Ensuring Sovereignty of the Republic of Belarus in the Field of Digital Development until 2030 under Resolution No 1074, the country's first coordinated strategy on digital sovereignty. It sets principles for national ICT control, faster rollout of new technology, and modernisation of government, economic and social services, folding AI in public administration into this wider push. As a foundational document it lays out a digital independence policy framework rather than detailed implementation steps.
Government of Luxembourg (tabled by Minister Elisabeth Margue in the Chamber of Deputies) — Draft Law No. 8476 enacts the EU AI Act (Regulation 2024/1689) at national level, part of AI regulation in Luxembourg, naming the CNPD lead market watchdog and sandbox operator. The CSSF and Commissariat aux Assurances cover financial and insurance oversight, with penalties and redress under EU AI Act implementation also fixed. As of July 2026 the bill still awaited committee approval, after opinions from the Council of State and other bodies.
Parliament of Albania (Kuvendi i Republikës së Shqipërisë) — Automated decision-making AI regulation in Albania runs through Law No. 124/2024, replacing the 2008 statute from 1 February 2025. Article 20 echoes GDPR Article 22: a purely machine-driven ruling, profiling included, cannot stand alone if it carries serious legal weight, absent human oversight. Article 53 extends this AI profiling safeguard to policing, blocking automated bias tied to sensitive data. Breach fines reach 2 billion Albanian lek or 4 percent of global turnover.
Pontifical Commission for Vatican City State — Vatican City State's Pontifical Commission adopted Decree N. DCCII, setting out Guidelines on Artificial Intelligence effective 1 January 2025. The rules shape AI governance in Vatican City State on data protection, non-discrimination and human oversight, with sector rules spanning health, heritage, courts and security. A five-member Commission on Artificial Intelligence within the Governorate will oversee this AI ethics framework.
Federal Chancellery (Bundeskanzleramt) — Version 2.0 of the Praxisleitfaden Digitale Verwaltung: KI, Ethik und Recht gives Austria's Federal Chancellery a working reference for AI ethics in public administration, covering how officials should plan, design, procure and deploy AI systems lawfully. New material folds in generative AI, quantum AI and EU AI Act duties, paired with a Checkliste für ethische KI in der Verwaltung 2.0 for practical use. A first edition preceded it in November 2023, before this update to AI in government administration tools.
Ministarstvo pravosuđa, uprave i digitalne transformacije (MPUDT) — Croatia now has seven bodies handling AI fundamental rights oversight. The Ministry of Justice, Public Administration and Digital Transformation listed them to the European Commission under Article 77(2) of the EU AI Act: the Ombudswoman, the Ombudswoman for Children, the Gender Equality Ombudswoman, the Ombudswoman for Persons with Disabilities, the Personal Data Protection Agency (AZOP), the State Electoral Commission and the Agency for Electronic Media. Each authority is enforcing rights protections against high-risk AI systems.
Ministry of Digital Government (Digitaliseringsministeriet) — Copenhagen commits DKK 62.5 million through 2027 to Denmark's AI strategy: four projects and three values for trustworthy machine intelligence. These cover a Digital Taskforce for AI, a centre on AI's societal impact, a Danish-language model platform, and open-source Danish text data — Denmark's national AI strategy.
High-Level Advisory Committee on Artificial Intelligence under the Prime Minister — Greece national AI strategy now exists in writing: "A Blueprint for Greece's AI Transformation", from the High-Level Advisory Committee on Artificial Intelligence under the Prime Minister, coordinated by the Special Secretariat of Foresight, formed November 2023. Six flagship programmes span government administration, schooling, research, technology and innovation. This is Greece AI policy framework in practice.
Austrian Federal Government (Federal Chancellery and Federal Ministry for Innovation, Mobility and Infrastructure) — The Federal Government's KI-Umsetzungsplan 2024 sets out 47 Austrian AI policy measures across all twelve federal ministries, extending AIM AT 2030 as an interim report. Eight priority areas include trustworthy AI, resilience and security, climate neutrality, AI literacy, digital administration and governance, favouring short- to medium-term action. This national AI strategy ties each measure to European values and the EU AI Act.
Ministry of Economic Development and Digitalisation — Moldova's Government staked its stance on AI regulation in a White Paper on Artificial Intelligence and Data Governance, drafted by the Ministry of Economic Development and Digitalisation to steer the nation's tech landscape. Non-binding, it signals intent, paving way for future measures unwritten. Its goal: ethical, inclusive, sustainable AI governance matching EU norms.
Federal Chancellery (Bundeskanzleramt) — Eight rules now shape AI use federal civil servants must follow in Austria's federal public service, set by the Federal Chancellery's KI-Guidelines. Staff keep oversight and final accountability, treating tools as merely supportive, restricted to data protection-compliant inputs, with generated content flagged openly. Outputs face screening for bias and ethics, and sustainability extends AI tools in public service beyond accuracy.
Belgian Data Protection Authority (DPA/GBA) — Belgium's Belgian Data Protection Authority released a 23-page brochure on 19 September 2024 under Belgium AI regulation, mapping how the EU AI Act meshes with the GDPR to guide AI compliance frameworks. Titled "Artificial Intelligence Systems and the GDPR - A Data Protection Perspective", it shows where core privacy duties like openness, fair processing, lawful basis, minimal data use, limited purpose and accountability now stretch to cover AI systems, plus fresh or heavier obligations placed on providers and deployers handling personal information.
Government of Andorra — A newly created body, the Data Intelligence Agency (AID), now leads Andorra AI governance under the Government of Andorra's Decree 327/2024. Its remit spans executing the national data and AI strategy, building a shared methodology, and drafting ethical and regulatory rules for artificial intelligence. AID also champions real-world projects blending data analytics with AI, shaping the country's Andorran AI governance strategy going forward.
Government of the Republic of Serbia — A decision by the Government of the Republic of Serbia, published in the Official Gazzette on 31 July 2024, gave the country its Council for Artificial Intelligence, a step shaping Serbia AI policy. Prof. Dr. Igor Peško leads the group, whose ranks pull together voices from public administration, universities and business to steer and check progress on the nation's AI development strategy. Beyond advising, members are charged with shaping and reviewing the strategy's action plan and pushing forward legislation and rules touching artificial intelligence.
Council of Ministers (Këshilli i Ministrave) — Albania's Council of Ministers made binding rules for AI use in public sector systems across the Republic of Albania through Decision No. 479, setting methodology and technical standards for artificial intelligence government systems. The central obligation requires human oversight built into each system, so people can review and correct outputs that turn out wrong or cause harm. Unlike voluntary guidance seen elsewhere, this measure carries legal force within Albania's public administration.
Government of the Czech Republic (via Ministry of Industry and Trade) — Czechia adopted its Czech Republic AI policy blueprint on 24 July 2024, naming targets to reach by 2030 across seven linked fields: research, teaching, skills-building, legal and ethical questions, security, commerce, and government operations. Officials behind the National Artificial Intelligence Strategy of the Czech Republic 2030 (NAIS) frame the document as a push to turn the country from a mere buyer of advanced AI systems into a builder of them, a shift also described as pursuing a national AI strategy for Czechia.
Agenzia per l'Italia Digitale (AgID) and Presidenza del Consiglio dei Ministri — AgID's Italy AI policy blueprint, the Italian Strategy for Artificial Intelligence 2024-2026, rests on four macro-areas: Research, Public Administration, Business and Training. Gianluigi Greco chaired the panel shaping it; the document adds a monitoring system plus regulatory analysis. It stands as Italy's leading Italian AI regulation reference through 2026.
Data Protection Commission — Ireland's Data Protection Commission issued 'AI, Large Language Models and Data Protection' on 18 July 2024, mapping GDPR and Irish law onto large language model training and use. The guidance addresses AI data protection concerns for organisations building generative AI tools, covering fairness, lawfulness, transparency, storage limitation, data subject rights and automated decision-making.
European Parliament and Council of the European Union — Spain AI regulation now runs on Regulation (EU) 2024/1689, adopted 13 June 2024 by the European Parliament and the Council and printed in the EU Official Journal, applying automatically with no domestic transposition step. Outright bans hit unacceptable AI practices, high-risk systems face fresh duties, and each member state must appoint a watchdog — AESIA got that job in Spain. A single bloc-wide text thus decides deployment rules under this risk-tiered EU AI Act.
Government of Romania — Romania AI policy for 2024-2027 rests on six priorities under the National Strategy in the Field of Artificial Intelligence (SN-IA): skills, infrastructure, research, technology transfer, wider societal uptake, and governance. Government Decision HG 832/2024, adopted 11 July, set up an Inter-ministerial Commission for the Coordination of the Implementation, drawing on 34 institutions. It doubles as national AI regulation, reaching government and society alike.
Ministry of Digital Transformation of Ukraine — Kyiv favours a phased path to AI regulation in Ukraine: self-regulation first, later a dedicated statute matching the EU AI Act. The Ministry of Digital Transformation's White Paper on Artificial Intelligence Regulation in Ukraine outlines a sandbox, conduct codes, labelling and sector advice, short of binding rules. Ukrainian AI law remains a proposal.
Malta Digital Innovation Authority — Malta launched the world's first national AI certification programme in October 2019 through the Malta Digital Innovation Authority's voluntary AI Innovative Technology Arrangement (AI ITA) scheme. Approval follows a risk-based assessment against published objectives on governance, security, auditability and data protection. The ITAS Certification Regulations (Subsidiary Legislation 591.01) fixed fees and legal mechanics for this AI compliance certification route, and the Technology Assurance Sandbox (MDIA-TAS) provides parallel testing for iterative development.
Federal Public Service Policy and Support (BOSA) / Federal Public Service Economy — A Royal Decree created a Steering Committee on Artificial Intelligence (Sturingscomité inzake artificiële intelligentie) within the Federal Public Service Policy and Support and the Federal Public Service Economy, advising the federal government on Belgium AI policy. The body is the sole federal contact point for AI governance, draws members from several ministries, regions, universities, the National Labour Council, BIPT, the Centre for Cybersecurity Belgium and the language communities, and convenes at least three times yearly.
Garante per la protezione dei dati personali — Italy's Data Protection Authority, the Garante per la protezione dei dati personali, issued guidance on web scraping generative AI training, giving preliminary direction on large-scale harvesting of personal information from online sources for this purpose. The note recommends protective measures for operators of Italian websites and platforms to guard against mass collection of personal data for AI models without consent.
Government of Spain, Ministry for Digital Transformation and Public Service — 1.5 billion euros: Spain has pledged that much behind the Artificial Intelligence Strategy 2024, adopted 14 May to refresh the older 2020 ENIA plan for faster-moving generative AI. One pillar builds capacity through supercomputing, the ALIA language models and talent pipelines; a second drives uptake via GobTechLab, Kit Consulting and Kit Digital; a third anchors ethics and oversight under the Spanish AI Supervisory Agency (AESIA). This is the backbone of Spain's AI strategy, ratified by the Council of Ministers, running through 2024-2025.
Seimas of the Republic of Lithuania — Lithuania's Seimas set eleven principles governing public-sector AI governance under Resolution No. XIV-2620: oversight, accountability, traceability, quality control, equality, non-abuse, transparency, protection of human interests. These AI deployment principles bind government bodies, with no stated enforcement mechanism.
Government of Andorra — Andorra's Government has issued a Code of Ethics on Artificial Intelligence, an official guide steering how ministries, companies and the public approach AI ethics code Andorra requires. Principles cover oversight by humans, openness, safeguarding data and answerability, mirroring benchmarks set by UNESCO, the OECD and the Council of Europe. Certain public-sector uses must follow it; everyone else may choose to apply this responsible AI governance framework.
Austrian Data Protection Authority (Datenschutzbehörde, DSB) — Austria's Datenschutzbehörde (DSB) claimed oversight of high-risk AI deployed in policing, courts and electoral processes, marking its debut FAQ on AI data protection compliance. Firms deploying such systems must meet GDPR's lawfulness, transparency, minimisation and accuracy tests, cross-referenced against the EU AI Act. Handling personal data through automated systems now falls under both frameworks simultaneously.
Cabinet Office — Procurement Policy Note 02/24 pushes AI in government procurement toward greater openness across Whitehall. Suppliers must flag any AI role in drafting bids, authorities must probe related risks, and contracts must bind data used to train AI models. Agencies, public bodies and central departments fall within this transparency in AI procurement drive.
Government of Flanders — Belgium's Flanders region has locked in five more years of funding under the Vlaams Beleidsplan Artificiële Intelligentie (Flanders AI Policy Plan), approved 22 March 2024 and covering 2024-2028. Money splits three ways: 14.185 million euros yearly for core research via the Flanders AI Research Programme, backing for firms adopting AI tools, plus ethics, legal and skills work run by the Knowledge Centre Data & Society and Flemish AI Academy. Total spend across the Belgium AI strategy lands near 34 million euros a year.
Digitaliseringsstyrelsen and Datatilsynet — Free GDPR guidance and AI Act risk classification are now open to Danish firms and public bodies through a sandbox jointly run by the Agency for Digital Government and the Data Protection Authority. This Danish AI regulation effort speeds movement from prototype to deployment, backing safer artificial intelligence adoption nationwide.
Federal Chancellery of Austria and Federal Ministry of Finance — Eleven experts now sit on Austria's AI Advisory Board (KI-Beirat), formed to steer federal AI policy on technical, ethical and societal grounds. Tasked with shaping the country's AI strategy, the panel also scans global trends for risks and openings facing Austria. RTR provides its back-office support and runs the linked AI Service Desk.
Office of the Ukrainian Parliament Commissioner for Human Rights — Ukraine's Ombudsman office, Ministry of Digital Transformation and the EU4DigitalUA project released joint guidance on AI data protection for public bodies and companies. Officials set ethical rules for artificial intelligence privacy compliance, stressing safeguards against rights abuses, as part of the country's Roadmap for AI Regulation.
Ministry of Economic Affairs and Communications — Estonia AI policy now centres on the AI and Data Action Plan 2024–2026, published by the Ministry of Economic Affairs and Communications to guide artificial intelligence across public agencies and businesses. An AI competence centre anchors the framework, which channels funding toward workforce skills, language technology and private-sector innovation while steering how government bodies deploy machine learning tools. This national AI strategy came into force on 8 February 2024 and remains active today.
Department for Science, Innovation and Technology — "Implementing the UK's AI Regulatory Principles: Guidance for Regulators" tells regulators how to apply five cross-sectoral principles within their own remits. UK AI regulation rests on safety, security, robustness, transparency, explainability, fairness, accountability, governance, contestability and redress. Each faced a 30 April 2024 deadline for a written plan on artificial intelligence oversight.
Austrian Parliament (Nationalrat) — Austria's Parliament passed a law reshaping AI regulation Austria-wide, amending the KommAustria-Gesetz and Telekommunikationsgesetz 2021. RTR now hosts the KI-Servicestelle (AI Service Desk), a single contact point guiding businesses and citizens through AI rules, while a KI-Beirat panel of eleven specialists advises the federal government on AI's ethical and societal fallout — an AI governance advisory board, not an enforcement body. BGBl. I Nr. 6/2024 carries the text, published 26 February 2024.
Department of Enterprise, Tourism and Employment — Ireland AI regulation now has an independent advisory voice: the AI Advisory Council held its opening session with Dr Patricia Scanlon chairing. Dara Calleary founded the panel to give ministers impartial specialist input, seeking public confidence in artificial intelligence built around trustworthiness and human needs. Members will also produce their own AI policy advice tracking new developments, dangers and possibilities.
The National Audit Office measured government AI adoption across 87 UK central government bodies for its March 2024 report. By autumn 2023, 37% had deployed AI and 70% were piloting or planning it.
Government of Spain, Ministry of Economic Affairs and Digital Transformation — Spain claims the European Union's first AI regulatory sandbox, created via Royal Decree 817/2023 (8 November 2023), where companies—chiefly SMEs and startups—pilot high-risk systems against draft EU AI Regulation demands ahead of enforcement. Working jointly with the European Commission, the project delivered 16 compliance guides through AESIA by December 2025, giving businesses practical AI compliance testing rather than theoretical guidance.
Datatilsynet (Danish Data Protection Authority) — Danish bodies building or rolling out AI systems must satisfy duties set by Datatilsynet: legal basis, informing those affected, and data protection impact assessments — new guidance on AI in public authorities. A companion survey tracked adoption across government nationwide.
Federal Public Service Policy and Support (BOSA) — Belgium's Raadgevend Comité Ethiek inzake Data en Artificiële Intelligentie (Advisory Committee on Ethics for Data and Artificial Intelligence), created by Royal Decree on 4 October 2023, shapes AI policy in government federally. Five experts on renewable five-year terms advise ministers on AI systems and data-driven technology, weighing rights, privacy, dignity and non-discrimination. This AI ethics in the public sector guidance is non-binding.
Government of Spain, Council of Ministers — Spanish AI regulation gained a dedicated authority when Spain's Council of Ministers signed Royal Decree 729/2023, forming AESIA (Spanish Agency for the Supervision of Artificial Intelligence) under a Statute effective 3 September 2023. As designated EU AI Act enforcer, AESIA certifies systems, fosters ethical conduct, connects with European institutions, and watches deployments spanning government and industry, cementing artificial intelligence governance beneath one roof.
Ministry of Research, Innovation and Digitalization — Ministerial ordinance OMCID 20484/2023 (Monitorul Oficial 382, 4 May 2023) sets the structure and operation of the Romanian Committee for Artificial Intelligence (CRIA). Government Memorandum 20D/31781/MN of November 2022 created CRIA as a Romania AI policy body under the Prime Minister's patronage. CRIA steers state and business AI projects, drafts rules and builds national AI governance capacity in research, development, education and deployment; MCID runs its technical secretariat.
Information Commissioner's Office — Fairness across the AI lifecycle sits at the centre of the ICO's revised guidance on AI and Data Protection, spelling out how UK GDPR duties reach accountability, transparency, lawfulness and accuracy in automated processing. The regulator sharpened its stance after industry pressed for clearer rules on AI data protection, backing safer adoption without weakening data protection in AI systems.
Autoriteit Persoonsgegevens (AP) — Netherlands' Autoriteit Persoonsgegevens (AP) hosts the new Dutch Directorate for Coordination of Algorithms (DCA), tasked with AI supervision for systems that threaten fundamental rights. Rather than replacing existing regulators, the DCA supports them through cross-agency risk analysis, closer collaboration between supervisory authorities, and input into new legal frameworks for algorithm oversight.
Council of Ministers / Federal Public Service Policy and Support (BOSA) — The Belgian Council of Ministers adopted the National Convergence Plan for the Development of Artificial Intelligence on 28 October 2022, launching SmartAI.Nation. This Belgium AI policy rests on nine strands: trustworthy AI, cybersecurity, competitiveness, a data-driven economy, healthcare uses, sustainable mobility, environmental safeguards, lifelong learning and upgraded public services. It remains the federal blueprint for AI governance frameworks.
Hellenic Parliament — Greece regulates artificial intelligence under Law 4961/2022, a national framework for emerging digital technologies and the country's core AI regulation Greece has enacted. Chapter B, Articles 3-14, mandates algorithmic impact assessments, transparency duties, and a public-sector registry of AI systems, alongside a Coordinating Committee for Artificial Intelligence overseeing this national AI framework.
Agência para a Modernização Administrativa (AMA) — Five evaluation pillars anchor GuIA Responsável (Responsible AI Guide), the framework Portugal's Agency for Administrative Modernisation (AMA) built for AI in public administration: accountability, transparency, explainability, fairness and ethics. A bundled scoring instrument lets teams assess systems against those pillars ahead of rollout. AMA aimed the guide at government bodies, yet academic institutions and private firms may equally adopt this AI governance framework for the public sector, since its scope is not confined to state agencies. Its debut coincided with the launch of AI4PA.
Department for Digital, Culture, Media and Sport (now Department for Science, Innovation and Technology) — This UK AI policy, the National AI Strategy, sets a decade-long plan to expand Britain's AI sector, spread adoption, and frame pro-innovation UK AI regulation. It unites funding, skills, oversight and infrastructure to build the UK as a top global AI power.
Austrian Federal Government (Federal Chancellery and Federal Ministry for Innovation, Mobility and Infrastructure) — Austria's national AI strategy, Artificial Intelligence Mission Austria 2030 (AIM AT 2030), rests on three pillars: trustworthy adoption rooted in European fundamental values, an AI research and innovation hub, and economic competitiveness. Health, climate, industry, education and public administration fall within scope. The document is Austria's central AI policy framework.
Swiss Federal Council — Switzerland's Federal Administration centralised AI in government via the Competence Network for Artificial Intelligence (CNAI/KNW KI), created by Swiss Federal Council decision, first hosted at the Federal Statistical Office. One unit covers advice, tooling and cross-agency coordination on government AI adoption, ending fragmented expertise. Six hubs split duties: governance and strategy, platforms, procurement, culture and training, law and regulation, cyber security, and data science. Since 1 February 2026 the Federal Chancellery, not the statistics office, runs the contact point.
Department of Enterprise, Tourism and Employment — Under Ireland AI policy, the Department of Enterprise, Tourism and Employment unveiled 'AI - Here for Good', the nation's inaugural National AI Strategy, targeting a top global position in trustworthy, human-centred artificial intelligence. Officials back an ethical, people-focused path for building, deploying and applying AI, meant to boost output, tackle social problems and upgrade public services via responsible AI adoption.
Government of Iceland (Prime Minister's Office) — Iceland's Prime Minister's Office built its Iceland AI policy, formally the national Policy on Artificial Intelligence, Stefna Íslands um gervigreind, on three pillars: benefit for everyone, a competitive workforce, and schooling fit for modern times. The plan binds government to ethical technology, adaptable jobs and protection of the Icelandic language online. It anchors later governance work, including the AI Action Plan 2025-2027, as the country's founding national AI governance framework.
Council of Ministers of the Republic of Bulgaria — Cabinet ministers backed the Concept for the Development of Artificial Intelligence in Bulgaria until 2030 via Protocol No. 72. Six linked strands make up this Bulgaria AI policy: scientific infrastructure, communications, lifelong learning, education, tech transfer, research, data rules, sector-based innovation, and an ethics framework matched to EU trustworthy-AI norms — an artificial intelligence strategy running to 31 December 2030.
Government of Spain, Ministry of Economic Affairs and Digital Transformation — Spain adopted ENIA, its National Artificial Intelligence Strategy, directing public funds toward AI in public administration and industry between 2021 and 2023. Sitting under Component 16 of the Recovery, Transformation and Resilience Plan, the scheme forms part of the broader España Digital 2025 push. It rests on six pillars: strengthening research, growing digital talent, scaling data infrastructure, embedding AI within value chains, extending it into government agencies, and establishing an ethical and regulatory framework. These figures come from the Spanish government's own stated commitment, not an independently audited result.
Ministry of Investments, Regional Development and Informatization of the Slovak Republic (MIRRI) — MIRRI created CERAI, the Permanent Commission for Ethics and Regulation of AI, as an independent expert advisory body on Slovakia AI regulation. The commission's 21 members weigh ethical, social and legal questions in regulating artificial intelligence across research, development, deployment and use, under a February 2021 statute.
Government of the Russian Federation — Government Directive No. 2129-r introduced Russia's Concept for the Development of Regulation of Relations in the Field of Artificial Intelligence and Robotics until 2024, charting the state's path on AI regulation and robotics policy. Officials laid out objectives, guiding principles and sector-by-sector duties spanning accountability rules, data handling regimes, technical benchmarks and insurance, alongside deploying AI across hospitals, transport networks, banking, smart urban systems and government agencies. This foundational text for AI governance rules continues in force under the Russian Federation's legal system.
President of the Russian Federation — Moscow gained a dedicated legal sandbox for testing artificial intelligence under Federal Law No. 123-FZ, which President Putin put into force to build conditions favourable for developing and rolling out AI systems within the Russian capital. Anonymised personal data may now be processed specifically to advance these AI regulation Moscow goals, alongside rules spelling out the trial's aims, guiding principles and operating procedure. Russia's State Duma cleared the measure on 14 April 2020, and the Federation Council gave final sign-off three days later, on 17 April 2020, completing the artificial intelligence experimental legal regime's legislative path.
Spanish Data Protection Agency (AEPD) — The Spanish Data Protection Agency (AEPD) issued a guide on Spain AI regulation titled 'Una aproximación para la adecuación al RGPD de tratamientos que incorporan Inteligencia Artificial' (An approach to GDPR compliance for processing that incorporates Artificial Intelligence). The guide covers privacy risks when regulating AI, plus key GDPR points for AI products and services: transparency, lawfulness of processing, risk management and automated decision-making.
Cabinet of Ministers of Latvia — Latvia's Cabinet of Ministers approved a national AI strategy via the report 'Par mākslīgā intelekta risinājumu attīstību' (Developing Artificial Intelligence Solutions), covering six areas: skills, trial projects, language technology, ethics, infrastructure and innovation. This AI adoption roadmap guides public and private uptake of artificial intelligence economy-wide.
Malta Digital Innovation Authority — Malta AI regulation rests on 'Towards Ethical and Trustworthy AI: Malta's Ethical AI Framework', from the Malta Digital Innovation Authority. Four pillars — autonomy, harm prevention, fairness, explicability — pair with 63 controls for trustworthy AI governance. It draws on EU High-Level Expert Group guidelines and OECD AI Principles, tying to Malta's national AI strategy, with oversight under a National Technology Ethics Committee at the MDIA.
President of the Russian Federation — President Vladimir Putin's Decree No. 490 launched Russia's National Strategy for the Development of Artificial Intelligence — a Russia AI strategy stretching to 2030 that covers research capacity, computing infrastructure, data access, skilled talent and uptake throughout government and industry. Decree No. 124 revised the plan on 15 February 2024, introducing fresh benchmarks for home-grown computing power, large-scale generative models, how widely organisations deploy AI, and trust rules governing public-sector systems. This places Russian national AI policy on a longer horizon than most comparable national plans.
Dutch government (Ministry of Economic Affairs and Climate Policy) — SAPAI, the Dutch government's Strategic Action Plan for Artificial Intelligence, sets over a hundred measures pairing public investment with joint work between industry and researchers. It funds entrepreneurs and scientists, builds digital infrastructure and data systems, and adds ethical and legal safeguards for citizens' rights under this national AI plan. The AI strategy targets economic and societal gains from AI for the country.
Malta Digital Innovation Authority and the Maltese Government — Malta now has a national AI strategy: 'The Ultimate AI Launchpad: A Strategy and Vision for Artificial Intelligence in Malta 2030', from the Maltese Government's Malta.AI Taskforce under the Malta Digital Innovation Authority. Three pillars anchor it — investment, public and private sector uptake — backed by education, ethics and infrastructure, plus 72 actions. Six pilots span traffic, health, tourism, utilities and other services, aiming for a top-ten global rank by 2030 — Malta AI policy in practice.
Government of the Czech Republic — The Committee for Artificial Intelligence (Výbor pro umělou inteligenci) oversees delivery of Czechia AI policy under the first National AI Strategy, tracking targets and endorsing action plans. Housed in the Government Council for the Information Society, within the Ministry of Industry and Trade, this AI governance body gathers several ministries, the Technology Agency, the Academy of Sciences and other stakeholders.
Ministry of Industry, Business and Financial Affairs (Erhvervsministeriet) — Four goals anchor Denmark's national artificial intelligence strategy: putting people first in AI ethics, building up Danish research capacity, helping companies expand through AI adoption, and pushing public services toward world-class delivery. Six ethical principles and 24 initiatives back the plan, funded with DKK 60 million across 2019 to 2027.
Federal Government of Germany (Bundesregierung / BMFTR) — German AI policy took formal shape when the Federal Cabinet approved the Federal Government's Artificial Intelligence Strategy, aiming to boost Germany's standing in AI research, development and application against international rivals. Revisions in December 2020 brought pandemic response and sustainability into scope, and the BMFTR later issued its AI Action Plan, mapping eleven fields of action and committing more than 1.6 billion euros in funding. This sequence marks Germany's shift from an initial artificial intelligence strategy toward binding investment pledges.
President of the Republic of Belarus — AI development Belarus gained legal grounding when the President's Decree No 8 took effect on 28 March 2018, licensing artificial intelligence, machine learning and neural network systems within the High Technology Park (HTP), plus autonomous vehicles and blockchain. HTP's tax and legal terms were extended to 2049. BelTA reports resident expertise now spans artificial intelligence in Belarus's digital economy, machine learning, Internet of Things and augmented reality — the root document centring AI in national digital strategy.
Ministry of Economic Affairs and Employment — Finland AI strategy, issued by the Ministry of Economic Affairs and Employment, aims to make the country a global leader in applying artificial intelligence across business, government services and society. Titled 'Finland's age of artificial intelligence', it sets out recommended actions for the Nordic nation to capitalise on AI-driven growth. The Finnish AI policy framework targets adoption across multiple sectors rather than one industry.
United KingdomThe UK government's Incubator for AI (i.AI), part of the Department for Science, Innovation and Technology, grouped over 50,000 Independent Water Commission consultation responses into policy themes in about two hours with its in-house 'Consult' tool, at £240 in computing. It says the job would otherwise take weeks; human experts needed 22 hours to check the output. DSIT estimates Consult could save about 75,000 days of manual analysis a year, roughly £20 million in staff time, across the 500-plus annual government consultations.
We research and keep track of applied AI globally. Government & Public Sector's record gained 75 entries in October. Your next dispatch will open with Government & Public Sector, then the most important news about how AI is being used, applied and regulated globally.
One edition a month. Unsubscribe from any of them.
Poland
Sweden
Italy
Lithuania
Spain
Norway
Ukraine
Switzerland
Russia
Latvia
Åland Islands
Malta
Austria
Bulgaria
Germany
Finland
Portugal
Netherlands
North Macedonia
Greece
Belgium
Croatia
Belarus
Serbia
France
Montenegro
Hungary
Czechia
Albania
Slovakia
Andorra
Estonia
Slovenia
Ireland
Gibraltar
Iceland
Bosnia and Herzegovina
Luxembourg
Liechtenstein
Isle of Man
Vatican City
Moldova
Romania
Guernsey
European Union
Jersey
Denmark
Faroe Islands
Monaco