Industrial Research Unit · Straits Institute for Applied AI
Applied AI in Government & Public Sector across North America
Every entry is dated and traced to a named source: 43 deployments, 22 regulations, 10 pieces of evidence of adoption and 5 pieces of evidence of impact.
Entries
80
Countries
2
Most recent
13 Oct 2026
More on Government & Public Sector across North America lands here as we find it.
Ottawa adds C$500 million to its Regional AI Initiative for business adoption
Innovation, Science and Economic Development Canada — Ottawa has put a further C$500 million over five years into the Regional Artificial Intelligence Initiative, lifting federal backing to C$700 million. Evan Solomon, Minister of Artificial Intelligence and Digital Innovation, said Canada's regional development agencies will direct funds to AI adoption by businesses and towards helping Canadian AI firms commercialise and grow, prioritising robotics, manufacturing, agriculture, transportation, natural resources, energy, health and life sciences. Since 2024 the scheme has backed 218 adoption or commercialisation projects.
Two-thirds of Canadian public servants now use AI, KPMG survey finds
KPMG in Canada — A KPMG in Canada poll of 519 federal, provincial and municipal staff puts AI use among public servants at 66 per cent, against 48 per cent the prior year. Ninety-four per cent deem guarding Canadian data and intellectual property pressing, while 76 per cent say purchasing rules steer buyers to homegrown AI suppliers whenever feasible.
Halton police use AI simulations to hire and train 911 call takers
Halton Regional Police Service — Halton Regional Police Service in Ontario has adopted AI training for 911 call takers and dispatchers, calling itself the first Canadian agency with this type of technology. GovWorx, a US firm, supplies mock emergency calls testing recruits' decision-making, communication and crisis-management ability; police say evaluation time per candidate has dropped 50 per cent. The same system serves 911 communicator training and will later assess hired staff.
McGill finds chatbots fed Quebec voters faulty election information
McGill University — Roughly 40% of Quebec voters asked AI chatbots about the election, and in the final three weeks 39% of ChatGPT's voting tips came from an amateur, error-strewn AI-built site, McGill University researchers told a federal parliamentary committee. Aengus Bridgman led the work; Taylor Owen of the Max Bell School of Public Policy testified.
Georgia's Secretary of State launches GABBY, a chatbot that answers residents' questions
Georgia Secretary of State — Georgia Secretary of State Brad Raffensperger has introduced GABBY (Georgia's Automated Bot, Built for You), a website chatbot on sos.ga.gov answering public queries solely from agency pages. The office expects the assistant to save Georgians time spent clicking through pages or ringing its call centre.
Palm Desert adopts a risk-tiered AI policy for city staff
City of Palm Desert — Palm Desert City Council unanimously adopted an AI policy grading each staff or official use of city-approved tools as low-, moderate- or high-risk. Higher-tier uses, such as handling sensitive city data, acting without prior human sign-off or creating legal or financial commitments, need approval from a department director or the city manager. The city's first artificial intelligence officer, named in March 2026, oversees it.
Boulder will route non-emergency police calls to an AI assistant
Boulder Police and Fire Communications Center — Boulder's Police and Fire Communications Center will hand its non-emergency line to ANET (Automated Non-Emergency Triage), an AI call assistant, soft-launching at the end of October 2026 for short spells several times weekly and going fully live by the end of November. Humans keep every 911 call: the automated system transfers anyone with a concern or unresolved request to a dispatcher, and staff follow up each case.
Phoenix police pilot AI that drafts reports from body-camera audio
Phoenix Police Department — Phoenix Police Department officers in the Mountain View Precinct are trialling AI report writing for 90 days, with software converting bodycam recordings into an opening version of each narrative. Patrol Bureau staff take part alongside the Family Investigations Bureau and the Assaults Unit within the Violent Crimes Bureau. Nothing enters official files until an officer has checked it, fixed errors and signed an attestation; the trial closes on 31 December 2026, followed by an evaluation.
Montgomery County launches Mo, an AI chatbot for voter questions
Montgomery County, Pennsylvania — Montgomery County, Pennsylvania, answers basic voter questions through Mo, an AI election chatbot drawing only on county voter services pages and mapping data, from polling places to mail ballots. Citibot of Charleston, South Carolina helped build the assistant, which runs 24/7 in over 70 languages. Mo won't endorse candidates or give voting advice.
Utah orders AI literacy training and AI use across state agencies
Office of the Governor of Utah — Utah Governor Spencer Cox has signed an executive order setting Utah AI policy across government. The Division of Technology Services designs AI literacy training for all agency staff, departments assess affordable AI processes and resident-facing automation, and the Department of Government Operations leads AI-led process upgrades plus safeguards against cyber threats and rogue AI agents. It builds on Cox's December 2025 pro-human AI initiative.
Carney launches a National Council on AI to advise on Canada's AI transformation
Office of the Prime Minister of Canada — Prime Minister Mark Carney created a 13-member National Council on Artificial Intelligence for independent federal AI policy advice. Yoshua Bengio, Ajay Agrawal and fellow members will offer national AI guidance on strategy updates, faster adoption, national champions, sovereign AI infrastructure and government transformation; Patrick Pichette of the new Digital Transformation Canada agency sits ex officio.
US launches America.gov, an AI front door to 29,000 government websites
US federal government (White House) — America.gov, the Trump administration's AI government chatbot built on Google's Gemini and xAI's Grok, handles US questions about passports, campsites and Social Security cards. Replies draw on roughly 29,000 official sites and currently redirect users; 2027 upgrades promise AI help with public services like finishing applications online.
Citrus County Sheriff's Office hands non-emergency calls to an AI assistant
Citrus County Sheriff's Office — Citrus County Sheriff's Office now sends the non-emergency and administrative lines, which logged over 126,000 calls in 2025, to AI call handling via Prepared by Axon. The Florida system transcribes live, turns other languages into English, poses follow-up questions and escalates suspected emergencies to a dispatcher. 911 traffic skips this automated call-taking and reaches a person.
Non-emergency call-takers and dispatchersyahoo.com
Manchester police put real-time translation into officers' body cameras
Manchester Police Department (Iowa) — Manchester Police Department in Iowa is issuing officers Axon body cameras offering real-time translation across up to 50 languages at roughly 95% accuracy. Non-English calls rose from 63 in 2024 to 112 in 2025; officers formerly relied on personal phone apps for live interpretation. Facing online data-protection worries, the department noted Axon's cloud audits every login and file access.
Canada and Manitoba fund applied-AI training for aerospace workers
Governments of Canada and Manitoba — Canada and Manitoba are funding the AeroTrace Workforce Pilot with $500,000, applied AI training for aerospace and manufacturing workers that also covers digital traceability and production data. Manitoba Aerospace and Defence will deliver the AI skills programme with mode40, a data firm, through the Canada-Manitoba Workforce Tariff Response Program.
Centralia requires supervisor sign-off for staff use of generative AI
City of Centralia, Missouri — The City of Centralia, Missouri, has adopted a generative AI policy with a sign-off form stating when staff may use AI. Employees need supervisor approval to apply AI to city business and remain answerable for checking all output, under the new AI rules for staff.
GenAI.mil logs more than 2 million users in one week
US Department of Defense — Over 2 million people used GenAI.mil within one week, Cameron Stanley, the Defense Department's Chief Digital and Artificial Intelligence Officer, reported of the prior week. Roughly 3 million personnel can reach this military generative AI service, offering Gemini for Government, ChatGPT Mil and Grok for Government. Gemini's Agent Designer, added in March, lets approved staff build defence AI tools.
Colorado Springs puts an AI agent on its non-emergency police line
Colorado Springs Police Department — Colorado Springs Police Department's Public Safety Communications Center has switched on Prepared by Axon to transcribe, translate and summarise calls and route non-emergency calls. S.A.R.A.H., an AI agent for non-emergency calls, now answers those lines first; trained call takers keep every 911 call. Automated call answering was tested from April 2026.
Tennessee launches Iris, an AI chatbot for 357,000 licensees
Tennessee Department of Commerce & Insurance — The Tennessee Department of Commerce & Insurance has rolled out Iris, an AI chatbot for professional licensing that answers round the clock for over 357,000 licensees in 26 regulatory programmes. Via the CORE system, in English or Spanish, the licensing assistant also looks up licensed individuals and businesses. A soft launch ran in August.
Licensing customer service representativefox17.com
Raleigh drops its AI customer-service avatar after a pilot
City of Raleigh — The City of Raleigh scrapped Daisy, an avatar-fronted AI virtual assistant for residents tested from April to July 2026 before a new city hall opens. Daisy missed a 90% accuracy target, often said information was unavailable, and residents shied away from the municipal chatbot. Trials covered English, French, Arabic and Spanish; findings went to the Smart Cities Connect Fall Conference.
New York opens a whistleblower route for AI company staff
Office of the New York State Attorney General — New York Attorney General Letitia James has told staff inside artificial intelligence companies to route AI whistleblower complaints — anonymously if preferred — through the Office of the Attorney General's secure portal. The alert leans on the Responsible AI Safety and Education Act, in force from January 1, 2027, the SHIELD Act and the federal Computer Fraud and Abuse Act, the scaffolding of this AI policy. James said technology's promise excuses nobody from laws protecting New Yorkers.
Missouri bars AI facial recognition from police plate readers
Office of the Governor of Missouri — Missouri Governor Mike Kehoe has signed Executive Order 26-18, a Missouri AI regulation for automated licence plate reader systems at state bodies and state-funded police. Most plate and vehicle data must be erased permanently within 30 days; vendors may not sell, share or commercialise Missouri-generated data; and the AI policy bars artificial intelligence facial recognition from ALPR systems. Kehoe cited constitutional and privacy concerns over indefinite storage of law-abiding citizens' movements.
Cambridge lets builders pre-check designs with Archistar AI
City of Cambridge — The City of Cambridge has switched on two AI tools for development approvals aimed at faster home construction. Archistar tests a proposed scheme against municipal requirements before anyone files, and the Cambridge Digital Twin renders the city in three dimensions so planners, developers and residents can see what is proposed. Since September 1 builders, architects and homeowners can put designs and requirements through this AI planning check.
TSA's Ace agent handles 100,000 traveller conversations a month
US Transportation Security Administration — Ace, the AI agent for airport security queries at the US Transportation Security Administration, built on Salesforce Public Sector Solutions, takes roughly 100,000 routine traveller conversations a month. 96% of those inquiries — packing liquids, clearing a checkpoint with a medical device — close without human escalation. Live since the summer travel surge, the travel AI assistant gives timely guidance.
GSA opens discounted ChatGPT to federal, state and tribal government
US General Services Administration — US General Services Administration has struck a OneGov agreement with OpenAI putting ChatGPT for government on consumption-based pricing. Running 27 months from 1 October 2026, the offer cuts token-based usage costs 50%, with no platform fee, minimum order or spend commitment. GSA counts roughly 3.5 million federal employees already reaching AI under earlier agreements. Federal, state, local and tribal bodies buy this government ChatGPT access directly, through resellers or cloud marketplaces.
California pilots AskCA assistant to guide residents to state services
California Office of Data and Innovation — California's Office of Data and Innovation is piloting AskCA, an AI-powered digital assistant giving residents a single front door to state and local services, grouped by life events such as looking for work, launching a business, family support and disaster recovery, Governor Gavin Newsom said. Beta testers can now sign up; Anthropic's Claude underpins its knowledge base, with security help from the California Department of Technology, following early trials with 140-plus fire-recovery leaders and Altadena wildfire survivors. CalHR will first run it live, powering AI job-skills matching on CalCareers from 30 September.
Canada launches $13 million AI literacy drive for students and teachers
Government of Canada — The Government of Canada has begun a $13 million National AI Literacy Initiative with the Alberta Machine Intelligence Institute (Amii), aiming at as many as one million post-secondary students and 50,000-plus educators from kindergarten to Grade 12, Minister of Artificial Intelligence and Digital Innovation Evan Solomon announced. Institutions may sign up from 21 September, when an educators' course opens its first chapter; a version for all Canadians comes later this year.
Mountain View pilots live AI translation at public meetings
City of Mountain View — The City of Mountain View, California, is piloting Wordly's AI translation at public meetings, letting residents follow in more than 60 languages; it reached a city council meeting on 8 September after debuting at an Environmental Planning Commission meeting a month earlier. The $58,367 pilot, including a $43,750 Wordly contract for 300 hours of live software translation, extends language access beyond the in-person Spanish and Chinese interpretation offered on request.
Dane County puts an AI assistant on its non-emergency line
Dane County — Dane County has put an AI chatbot for non-emergency calls, the Aurelian Virtual Assistant or AVA, on the 608-255-2345 line, fielding information requests, department referrals and routine questions since Aug. 18. A one-year contract worth $155,000 covers the software, 90% of that from grants and 10% from county funds. Officials stress that no human communicator or dispatcher is being replaced: the AI call handler exists to leave staff free for emergencies.
Missouri opens free Google AI training to educators statewide
Missouri Department of Elementary and Secondary Education — Missouri's Department of Elementary and Secondary Education and its Department of Higher Education and Workforce Development have opened free Google AI training for educators across a state holding nearly 100,000 educators and 1.1 million students; adoption stays at the discretion of districts and charter schools. The bundle: Google Career Certificates and Google AI courses, Gemini for Education, Gemini Notebook and the Google AI Educator Series. Uses span lesson planning, grading, customised learning materials, research and an AI literacy curriculum. Registering with Grow with Google secures a free licence, a wide slice of state AI skilling.
Canada's refugee board bans AI-written evidence in its hearings
Immigration and Refugee Board of Canada — The Immigration and Refugee Board of Canada now applies a practice notice on AI in refugee hearings across all four divisions. No party may have AI create or materially alter personal evidence, whether witness statements, affidavits or basis of claim narratives. Under these tribunal AI rules, filers answer for submissions, must verify AI-generated content and must disclose AI transcription or translation.
Quebec courts bar judges from using generative AI to decide cases
Quebec courts — The Quebec Court of Appeal, Superior Court, Court of Quebec and municipal courts jointly set generative AI rules for judges. Substantive legal reasoning, characterising or analysing facts, weighing evidence or credibility, deciding outcomes and drafting those passages stay off-limits; correcting, revising, rephrasing or translating a judge's own text, some administrative work and searching specified sources are allowed. These judicial AI guidelines followed a March report of a Quebec ruling citing nonexistent case law.
Nebraska takes its DMV AI assistant statewide after pilot
State of Nebraska — Nebraska has chosen Tyler Technologies' Resident AI Assistant for statewide use after a May–August pilot at the Department of Motor Vehicles. During the pilot, the state says, it answered over 88,000 questions from 38,000-plus users, made more than 3,400 referrals to state online services and cut call volume by 20 per cent in some areas, with nearly a fifth of interactions after hours. It launches on Nebraska.gov as Nebraska Navigator before reaching participating agencies, and the DMV is widening its use of Tyler's document automation for motor carrier operations; deputy director Mitch Greenwall said the pilot showed AI improves resident experience while easing pressure on agency resources.
The Pentagon activates ChatGPT across a 3-million workforce
US Dept of Defense — The US Department of Defense activated ChatGPT Mil, OpenAI's model, on its GenAI.mil platform, accredited for controlled unclassified information at Impact Level 5, alongside Google Gemini and Grok for Government. The platform serves document-heavy planning, policy, logistics and administration across a workforce of more than 3 million. The Pentagon reports GenAI.mil already has over 1.7 million unique users, and the Navy and Marine Corps require it enterprise-wide.
Toronto police put AI plate readers across the force
Toronto Police Service — Toronto Police Service has deployed AI-powered cameras that read licence plates plus vehicle makes and models, which the service rates highly effective against crime. The system logs where and when vehicles are spotted and raises alerts officers can act on. Privacy researchers warn of possible misuse and mass surveillance while automatic plate recognition spreads through Canadian forces.
National Archives says AI use alone does not create federal records
US National Archives and Records Administration — The US National Archives and Records Administration's August memo (AC-11-2026) tells agencies that using AI does not, 'in and of itself', produce federal records. Whether AI material must be kept turns on factors like an agency relying on it for decisions, using it for official business, circulating it or building it into an agency system; NARA recommends formal AI policies.
CIA uses GPT-4.1 to compile open-source World Cup intelligence reports
US CIA — The US CIA's Open Source Enterprise used GPT-4.1 to help compile open-source intelligence reports on the 2026 FIFA World Cup, sharpening article selection, translating titles and summarising host-country coverage. Drawn from unclassified, publicly sourced material and checked by CIA analysts at every stage, the reports went to local law-enforcement partners via fusion centres such as the Washington State Fusion Center.
US Veterans Affairs launches a benefits help-desk chatbot in beta
US Dept of Veterans Affairs — The US Department of Veterans Affairs is beta-testing an LLM-driven help-desk chatbot on its 'Contact Us' page, letting veterans ask conversationally about benefits and services. It steers users to resources but offers no route to a person, is unsuitable for emergencies or medical questions and gathers no personally identifiable information, the VA says.
Calgary scores 17,000 intersections with AI to rank pedestrian risk
City of Calgary — The City of Calgary has used an AI platform to score over 17,000 intersections lacking signals or flashing beacons from zero to 100 for pedestrian risk, publishing the 50 riskiest. All scored 93 or more, led by Martindale Boulevard and Martindale Gate N.E. (99.02). Staff recommend kerb extensions, rapid flashing beacons and similar fixes at 26 of them, costing roughly $6 million, with design in 2027 if council approves funding in November's budget talks. Ward 5 councillor Raj Dhaliwal backs data-driven decisions but wants the model's reasoning explained; last year 15 people died on Calgary streets.
California opens its in-house AI assistant to state departments
California Department of Technology — Poppy, the California Department of Technology's generative AI tool, routes work to Claude, Gemini and ChatGPT with data kept in state systems, and now serves about 1,000 staff in 24 departments. Built on existing budget after a 2025 pilot (2,700-plus people, about 70 departments), it charges departments $7,500–$35,000 to join plus per-user fees. Lucy Andrews of the Department of Water Resources finds the tasks it absorbs often dull and slow. Users remain well below one per cent of state staff.
Defense Innovation Unit begins testing an AI military-logistics tool
US DoD Defense Innovation Unit — The US Department of Defense's Defense Innovation Unit began prototype testing of the Joint Sustainment Decision Tool, AI for predictive military logistics, in August 2026, with tests running to March 2027. It is meant to foresee and reduce risks across military sustainment and the supply chain.
IRS Policy for Artificial Intelligence Governance (IRM 10.24.1)
Internal Revenue Service — Internal Revenue Service governance for AI now runs through a revised IRM 10.24.1, the IRS Policy for Artificial Intelligence Governance, driving IRS-wide AI governance in tax administration toward faster, trustworthy builds and rollouts. The Chief Data and Analytics Officer, now Responsible AI Official, must apply stricter rules to high-impact systems and maintain an AI use case inventory. The framework aligns with OMB M-25-21 and other federal mandates.
US Army HR Command deploys Agentforce AI for 9.2 million people
US Army Human Resources Command — The US Army Human Resources Command rolled out Salesforce's Agentforce Public Sector platform for round-the-clock HR support to about 9.2 million military families, veterans, soldiers and civilian employees, via HRC's Digital Front Door personnel-records and HR-services platform. HRC says it is the first US military body on the platform since its Impact Level 5 authorisation, letting AI agents handle controlled unclassified information.
New Orleans puts AI on live 911 calls to triage incidents
New Orleans — Orleans Parish Communication District — The Orleans Parish Communication District has put Carbyne's AI on live 911 calls to triage incidents and screen out duplicates, notably from rush-hour crashes, making New Orleans the first major US city to disclose one. Carbyne says the AI triaged over 3,500 events in 90 days, cut redundant calls over 30% and sped responses by up to 40 seconds, figures not independently evaluated. It learned from real 911 recordings, local street names and speech patterns.
US Defense Department clears Salesforce's Agentforce at Impact Level 5
US Dept of Defense — The US Department of Defense has authorised Salesforce's Agentforce 360 at Impact Level 5, clearing the agentic AI system to handle controlled unclassified information and some national security data. Offered via Salesforce's Missionforce National Security suite, it will support command insights, onboarding, logistics and administrative workflows on sensitive unclassified missions, the company says.
Alberta turns AI agents on 466 million lines of government code
Government of Alberta — Alberta's Ministry of Technology and Innovation has built and deployed AI agents on Anthropic's Claude models to review more than 466 million lines of government code, after 18 months of tool-building. The review gave the province its first full picture of system health, and it plans to use the agents to replace 185 ageing systems with 16 applications it fully owns. Its 21 technical papers are open for other governments to reuse.
Toronto to add an AI chatbot to its redesigned 311 site
City of Toronto — The City of Toronto says it will launch its own AI chatbot for general public enquiries on its 311 site, leaving phone agents for harder issues. Part of a redesigned 311 website due by the end of 2026, it could explain how to file a request or how long a pothole repair takes. The city says 311 gets 1.5 million calls a year, about 60 per cent general enquiries.
Innovation, Science and Economic Development Canada — In Toronto on 4 June 2026, Prime Minister Mark Carney rolled out AI for All, a national AI strategy setting Canada's five-year course on artificial intelligence. Officials expect CAD 200 billion added to the economy and 250,000 fresh AI-related roles, financed through incoming laws and programmes meant to keep rollout responsible. Public confidence, opportunity and Canadian sovereignty underpin the wider push toward AI adoption.
12,000 Ontario civil servants used unapproved AI websites
Office of the Auditor General of Ontario — About 12,000 Ontario Public Service employees browsed close to 400 GenAI-related sites across five months, an audit by Ontario's Auditor General discovered, drawing on Microsoft Defender logs to reveal shadow AI in government spreading through Canada. Microsoft Copilot Chat alone holds security approval for OPS work, yet captured only six percent of that browsing — a clear sign of unapproved AI websites in daily use. Auditor Shelley Spence's review also pointed to persistent holes in the province's wider artificial intelligence strategy.
Burlington opens AI plan checking to building permit applicants
City of Burlington — eCheck lets applicants for a Burlington building permit pre-test designs free before formal submission. The Canadian City's AI building permit review tool checks proposals against zoning bylaws and the Ontario Building Code, covering ceiling heights and room sizes plus fire protection, setbacks, height, floor ratios, landscaping and parking. Mayor Marianne Meed Ward said Burlington is first in Canada to apply automated building code assessment pre-submission, ahead of other municipalities on early compliance checks.
Pennsylvania widens ChatGPT Enterprise to 3,000 state employees
Commonwealth of Pennsylvania — Over 3,000 Commonwealth employees across 35 agencies in Pennsylvania, United States, now have access to ChatGPT Enterprise under an expansion by the Shapiro Administration, part of a wider state government AI deployment, with 6,500 more staff required to complete training on safe and responsible AI use before gaining entry. This follows a 12-month pilot in which 175 workers spanning 14 agencies reported cutting an average of 95 minutes from their daily workload while drafting messages, condensing documents, researching topics and generating ideas, an early sign of AI in state government paying off. Officials note human oversight stays central to the effort, and flag that the 95-minute saving is a self-reported figure, not independently verified.
Michigan reads every SNAP case with AI before payment
Michigan Department of Health and Human Services — Michigan's Department of Health and Human Services now scans every SNAP case with an AI case-reading tool before payment, built on Google Vertex AI with optical character recognition reading pay stubs and other documents. David Knezek told the Senate Appropriations Subcommittee on DHHS that this AI benefits eligibility checking, an automated eligibility review, flags cases most likely to error, though staff still check each flag. It follows MiDAS, which wrongly charged over 40,000 people in the United States with misrepresentation within two years; a later audit of 22,000 cases found 93% involved no actual fraud.
Massachusetts gives 40,000 state employees a ChatGPT assistant
Commonwealth of Massachusetts — Massachusetts governor Maura Healey announced that the state's 40,000 executive-branch employees will get a ChatGPT-powered AI assistant, deployed with OpenAI after competitive procurement. Rollout is phased, beginning with the Executive Office of Technology Services and Security, then other agencies, and the state says the tool keeps prompts in a walled-off environment where they cannot train public models. Other states have run limited ChatGPT pilots; this is the first across a whole branch of state government.
Los Angeles courts give judges an AI clerk for draft rulings
Los Angeles County Superior Court — Los Angeles County Superior Court is piloting Learned Hand, an AI tool that works as a judges' clerk, distilling long filings, writing research memos and drafting orders. The court's contract, worth roughly $314,000, maps out testing in criminal, family and probate divisions, though judges so far mostly use it on civil motions; Riverside County's court is trialling it under a $10,000 agreement. Learned Hand combines Anthropic, OpenAI and Google models, and says it checks bias and accuracy but has published no results.
Federal Court of Appeal gives registry staff an AI assistant
Federal Court of Appeal — Canada's Federal Court of Appeal built Amicus, an AI assistant for court registry staff answering procedural and operational questions. It draws sourced, verified material from a central database, so matching queries yield matching answers rather than results that vary by employee. Feedback has been strongly favourable; the goal is consistent, national service quality for internal teams and external litigants alike, functioning as a broader AI registry assistant.
Louisville Metro Government — Louisville Metro Government's new AI housing permit process pilot, starting this January, picks Govstream.ai to scan zoning codes, geographic records and past filings for recurring delays. Officials expect faster reviews, clearer visibility and earlier guidance for applicants and staff alike. Outcomes from the AI-driven building permit scheme go public once the spring trial ends.
Connecticut bars police from using AI to draft criminal reports
Connecticut's Chief State's Attorney imposed a statewide moratorium in 2026 on police using AI tools to draft, write or narrate criminal reports, warning that officers would swear under oath to computer-generated text. The King County, Washington prosecutor's office takes the same stance.
Windsor approves AI chatbot pilot for provincial offences
City of Windsor — Windsor's Provincial Offences Administration office will launch a voice-enabled AI chatbot for municipal government, answering resident calls over a six-month trial. Officials forecast it absorbing about 30 per cent of clerks' duties, with its $57,600 yearly cost offset in 10 months via staffing savings from this virtual assistant deployment.
Quebec publishes AI framework for all public bodies
Government of Quebec — Quebec's Government has ended a March 2025 ban on public bodies using AI virtual assistants, alongside two new papers on AI regulation public sector rules. One directive sets requirements for generative AI governance, risk management and data protection; a second updates guiding principles for trustworthy use. Together they frame how artificial intelligence deployment should proceed across every provincial public body.
Canada launches first public register of AI uses in federal government
Government of Canada — Canada's Treasury Board President Shafqat Ali unveiled a first-of-its-kind AI Register, giving Canadians AI transparency in government over how federal agencies use artificial intelligence. Hundreds of entries span automated decision tools and machine-generated policy drafting, letting residents track where departments test or deploy these systems. The launch delivers on a pledge inside the federal AI Strategy for the Public Service 2025-2027, promoting broader visibility into government AI deployment nationwide.
Ontario gives 15,000 civil servants Microsoft Copilot in AI rollout
Government of Ontario — Over 15,000 Ontario civil servants use Microsoft Copilot weekly, Canada's largest known AI rollout in government, per internal figures. Common uses include tracking coverage and writing releases, while several hundred workers get deeper access through an advanced tier. It marks a growing embrace of generative AI in the public sector.
Canada pilots GCtranslate AI tool in six federal departments
Government of Canada — Six federal agencies, PSPC, the Privy Council Office, Finance Canada and the RCMP among them, are testing GCtranslate, an AI translation in government pilot from Canada's Translation Bureau. Officials envision expanding the system across the entire public service eventually. Trained linguists assessed the AI translation tool for government use beforehand, after it handled more than 60 million words in prior trials.
Federal Court of Canada publishes AI principles barring automated rulings
Federal Court of Canada — Canada's Federal Court will withhold automated decision-making tools from judgments until public consultation occurs, per Interim Principles and Guidelines steering judicial AI governance, which covers automated tools in court rulings. Staff may still tap such systems for translation and clerical duties, so long as humans verify every output.
Federal Court interim principles and guidelines on AI
Federal Court of Canada — Canada's Federal Court bars judicial AI decision-making in judgments and orders without prior public consultation, under its Interim Principles and Guidelines on the Court's Use of Artificial Intelligence. Seven principles govern: accountability, rights, non-discrimination, accuracy, transparency, cybersecurity, human oversight. The standard is human-in-the-loop court decision-making, not automation.
Transparency in Frontier Artificial Intelligence Act (SB 53)
State of California — California Governor Gavin Newsom enacted Senate Bill 53 (Transparency in Frontier Artificial Intelligence Act), frontier AI regulation obliging major developers to post a safety framework. These advanced-model rules create a CalCompute consortium for public computing, route critical safety incidents to the Office of Emergency Services and shield whistleblowers exposing health and safety risks.
Half of Canadian public servants using AI rely on public tools, survey finds
KPMG in Canada — AI adoption in government stays low across Canada: KPMG in Canada polled 349 Canadian public servants, finding just 22 per cent of public sector bodies have formally embraced AI. Yet 48 per cent of staff use it unofficially, half via open tools such as ChatGPT instead of vetted systems — fuelling accuracy, privacy and security exposure. Weak literacy compounds this: 85 per cent call public sector AI knowledge poor, and 69 per cent confirm real unease over generative AI tools in the public sector.
The White House — "Winning the Race: America's AI Action Plan" spans three pillars — innovation, AI infrastructure, and diplomacy and security — via over 90 federal policy actions. A January order prompted this White House blueprint, now the United States' national US AI policy, replacing scattered US AI regulation with one roadmap.
RCMP pilots Axon AI to draft police reports from body camera audio
Royal Canadian Mounted Police — Eight British Columbia detachments plus two more in Alberta are piloting AI police report writing for the RCMP, feeding body-worn camera sound into Axon's Draft One to generate draft incident reports automatically. A $200,000 ceiling covers the 12-month trial, though human review stays compulsory — each AI-assisted incident report needs at least 10% altered by an officer before approval. The system draws on audio alone, excluding video or facial-recognition data entirely.
Mount Royal launches Villemo conversational robot on municipal website
Town of Mount Royal — The Town of Mount Royal's website now runs Villemo, a conversational robot municipal website tool answering resident questions on demand. It reviews queries so officials spot recurring concerns and fine-tune services, storing no personal data, an AI chatbot for municipal services built with that privacy safeguard.
FDA launches Elsa generative AI tool across the agency
US Food and Drug Administration — Elsa, a generative AI tool FDA employees now use daily, handles scientific reviews, operational tasks and clinical evaluations inside the Food and Drug Administration. Running on a secure GovCloud setup, the system condenses paperwork, spots adverse events, drafts label comparisons and builds database code. It arrived early, beating a 30 June deadline after a strong pilot, launching a wider AI in drug regulation and clinical evaluation effort across all 11 FDA centres.
Markham launches Millie as face of upgraded AI chatbot 3.0
City of Markham — Canada's first municipal mascot, Millie the butterfly, now represents Markham's AI chatbot for municipal government use. Built on Comm100, OpenAI and IBM Watson under the Chatbot 3.0 upgrade, this AI virtual assistant for city residents answers queries 24/7 via council website content.
Accelerating Federal Use of AI through Innovation, Governance and Public Trust (M-25-21)
Office of Management and Budget — Memorandum M-25-21 pushes federal use of AI across US agencies, replacing M-24-10 with fresh demands: a named chief AI officer, an AI governance board, and risk controls on high-impact systems. Innovation now sits alongside stronger federal AI governance and public trust as core aims.
Driving Efficient Acquisition of Artificial Intelligence in Government (M-25-22)
Office of Management and Budget — Memorandum M-25-22 from the Office of Management and Budget replaces Biden-era M-24-18 as United States AI policy for executive agency tenders released after October 1, 2025. Buyers must assess whether a purchase counts as high-impact AI. These AI buying rules demand clauses on performance validation, interoperability and government data rights, and ban training commercial AI with non-public government data.
Virginia signs law expanding defamation rules to AI-generated content
Commonwealth of Virginia — Virginia now treats fabricated or altered artificial-intelligence material as grounds for defamation claims, under a fresh AI content regulation named HB 2124, the Synthetic Digital Content Act. Anyone who fraudulently misuses such content commits a Class 1 misdemeanour, and people portrayed without consent can pursue civil lawsuits against the wrongdoer. Virginia's Attorney General must convene a group to examine how statutes against AI-generated deepfakes get enforced.
SSA completes nationwide rollout of AI hearing transcription system
Social Security Administration — HeaRT (Hearing Recording and Transcriptions), SSA's AI hearing transcription system, now covers every hearing site nationwide, producing records for about 500,000 disability hearing customers yearly. Old hardware is retired; SSA projects $5 million in yearly savings on transcription and maintenance. This marks SSA's shift toward automated hearing-record technology.
Canada launches first AI strategy for the federal public service
Treasury Board of Canada — Canada's Treasury Board has unveiled its AI strategy for federal public service work, a 2025-2027 plan from Ginette Petitpas Taylor at the University of Waterloo. The strategy sets up a centre of AI expertise, mandates security and responsible-use rules, builds talent pipelines, and pledges transparency on government AI adoption plans.
Pennsylvania state staff saved 95 minutes a day using ChatGPT
Pennsylvania employees in a yearlong state-government ChatGPT Enterprise pilot, ended March 2025, reported saving 95 minutes a day on average. The 175 participants in 14 agencies self-selected and self-reported.
CAL FIRE runs 1,263 AI camera sensors to detect wildfire smoke
CAL FIRE and UC San Diego run ALERTCalifornia, a network of 1,263 AI-enabled camera sensors that auto-detect smoke and alert dispatchers in all 21 CAL FIRE command centres. In 2025 it helped inform responses to about 3,600 wildfires, flagging over half before any 911 call, says ALERTCalifornia director Neal Driscoll; no independent evaluation exists.
Law enforcement is the biggest AI use category at US Homeland Security
Law enforcement was the largest AI use category at the US Department of Homeland Security in 2025: 86 of 238 reported AI use cases, including 49 at Customs and Border Protection and 29 at Immigration and Customs Enforcement. Of those, 35 were classed high-impact.
US federal agencies report 3,611 AI use cases in 2025
The US federal government's 2025 AI use case inventory lists 3,611 individually-reported AI use cases across 56 agency submissions, 445 of them high-impact. The Department of Justice reported 314 (114 high-impact) and the Department of Homeland Security 238 (55 high-impact).
Your essential monthly update on Applied AI, with Government & Public Sector specially tracked
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.
Canada
United States