Questions & answers

Software that finds patterns in data and produces an output such as a transcript, a draft report, a match score or a map. It does not understand a case the way an investigator does. See AI Explained.

Publicly reported uses include report drafting from body-camera audio (RCMP and Fredericton pilots), transcription and translation (Edmonton, Calgary), image comparison (York and Peel) and crime analytics (Lethbridge). [Blue Line / Canadian Press, April 2025] [CBC / Canadian Press, June 2026] Full list: Use Cases.

Some have, as publicly reported. York and Peel Regional Police launched a shared system in May 2024 for comparing crime-scene images with mug shots, and Toronto officers used Clearview AI in 84 investigations before the chief ordered them to stop. [Blue Line / Canadian Press, April 2025] [CBC News]

As claimed by services: less time on paperwork and manual comparison, and less officer exposure to abusive material. The RCMP says it is piloting Draft One to see whether it can reduce the time and resources needed for reports. [CBC / Canadian Press, June 2026] These are service claims; we have not seen independent evaluations. [Blue Line / Canadian Press, April 2025]

The sources we reviewed describe tools that assist people, not replace them. The RCMP said Draft One is not intended to remove the need to write reports entirely. [CBC / Canadian Press, June 2026] The Toronto policy says members remain accountable for decisions made with AI assistance. [TPSB policy] Labour concerns are also being raised more broadly; for example, Ontario public-sector unions are pushing for workplace AI safeguards. [Global News, Oct 2026]

There is no single law on police AI. Privacy law, the Charter and human rights law apply, but privacy commissioners say the legislative context for facial recognition is insufficient. [Joint guidance, May 2022] The Toronto Police Service Board policy says no current statutes or regulations fully govern AI use in Ontario or Canada. [TPSB policy] Whether a specific use is lawful depends on the tool, the data and the authority relied on; ask your legal counsel. See Law & Rules.

Not one that covers policing. As reported in April 2025, the federal Artificial Intelligence and Data Act died when Parliament was prorogued on 6 January 2025 and applied only to the private sector. Ontario was first to adopt legislation on government use of AI, though the Law Commission of Ontario said it is unclear how far it covers police. [Blue Line / Canadian Press, April 2025] We have not checked later developments.

Examples: the Toronto Police Service Board policy [TPSB policy]; the joint privacy-authority guidance on facial recognition [Joint guidance, May 2022]; and draft CACP principles for responsible AI use, on which the CACP asked members for input in October 2026. [CACP news alert, Oct 2026] [CACP principles (FR), July 2026] Details: Law & Rules.

Clearview AI built a facial recognition database from images collected online, and the RCMP and other Canadian police used it. Canada's privacy authorities found Clearview's practices unlawful in 2021, and the Privacy Commissioner found the RCMP contravened the Privacy Act. [Joint findings, 2021] [OPC news release, June 2021] Timeline: Law & Rules.

Canadian guidance says a person should review matches before any decision to detain, investigate or charge. [OPC, 2021] A York Regional Police spokesperson said a match alone would not be grounds for an arrest. [Blue Line / Canadian Press, April 2025]

The sources we reviewed do not point to a Canadian law that specifically authorizes or bans it. The Canadian Civil Liberties Association has objected to some techniques, and Citizen Lab argues algorithmic policing can engage Charter rights and embed bias. [Blue Line / Canadian Press, April 2025] [Citizen Lab, 2020] These are advocacy and research positions, not court rulings.

Some pilots do. Axon's Draft One turns body-camera audio into a draft that an officer reviews; the RCMP says officers must change at least 10 per cent of what it produces. [CBC / Canadian Press, June 2026] A US supervisors' guide warns of "polished inaccuracy". [Police1]

Unknown. The RCMP said that as of June 2026 no incident report from its Draft One pilot had reached a courtroom, and legal experts quoted by CBC raised questions about accuracy and about which parts were written by the officer. [CBC / Canadian Press, June 2026] The Toronto policy requires the possible influence of AI to be included in disclosure to the Crown where it may lead to charges. [TPSB policy] Talk to your Crown and counsel about local practice.

That depends on each service's policy. A Police1 article for supervisors (US) advises services to define approved tools and what data may be entered, and to say clearly if AI-assisted writing is not allowed. [Police1] If unsure, ask your supervisor or privacy office before using one.

The Toronto Police Service Board policy requires the possible influence of AI to be included in disclosure to the Crown where it may lead to charges, and says reviews should be public to the greatest degree possible. [TPSB policy] That policy applies to Toronto; ask your own service about its rules.

Regulators highlight collection without lawful authority, accuracy and bias, retention, and lack of transparency. [Joint guidance, May 2022] For tools that send data to vendors, ask where data is stored and whether it trains vendor models; see the vendor questions.

Some tools can be. The OPC notes facial recognition has a propensity to misidentify certain gender and racial groups. [OPC, 2021] Citizen Lab argues algorithmic policing can embed systemic bias. [Citizen Lab, 2020] Testing for accuracy and bias before and during use is a recurring recommendation.

The Toronto policy says members are accountable for all the decisions they make, "including those made with the assistance of AI technology". [TPSB policy] The OPC guidance calls for human review of matches. [OPC, 2021]

Policy first, risk review, a named accountable owner, training, and a plan for public transparency are common themes in the sources. [TPSB policy] [Joint guidance, May 2022] See the adoption checklist.

Start with where the training data came from and whether it was lawfully collected in Canada, what independent testing exists, where your data is stored and whether it trains vendor models, and what audit logs you can produce. [OPC, 2021] Full list: vendor questions.

Yes. We offer solutions in digital notes, wellness, work process review, training systems and systems integration consulting. See Solutions or write to [email protected].

Write to [email protected]. See About & editorial policy for how we handle corrections.