A public-interest observatory documenting AI incidents and hazards in Canada. Structured evidence for prevention and accountability.
The White House TikTok account posted AI-altered video of a U.S. Olympic hockey player and Ottawa Senators captain, fabricating anti-Canadian slurs. The video received over 11 million views.
A 19-year-old used AI face-swap to put Ghislaine Maxwell's face on a Quebec City pedestrian; the video went viral with 7M views, leading to widespread conspiracy theories.
OpenAI's safety systems flagged and banned a ChatGPT account for violent content in June 2025. The account holder carried out a mass shooting in Tumbler Ridge, BC in February 2026. OpenAI had not reported the flagged account to law enforcement. The incident prompted federal calls for mandatory AI safety reporting requirements.
Toronto Police said AI scams 'took off like a rocket'; Competition Bureau warned of AI government impersonators; Toronto fraud losses hit $433M in 2025 with national losses reaching $704M.
Google's AI Overview generated false criminal accusations against a Canadian musician, leading to a concert cancellation.
Edmonton Police launched the world's first facial recognition body camera pilot in December 2025, scanning faces against a watch list of 6,341 people in silent mode without real-time field alerts. EPS stated regulation requires submission of a privacy assessment but not prior approval; Alberta's Privacy Commissioner rejected this interpretation.
Canada's only AI bill (AIDA) lapsed when Parliament was prorogued in January 2025. No replacement has been tabled. The government has adopted a 'light, tight, right' approach. 85% of Canadians support AI regulation; 92% are unaware of any existing AI laws.
Multiple frontier AI models have demonstrated deceptive and self-preserving behavior in controlled evaluations. Mila co-authored foundational research. These models are available to millions of Canadians. No Canadian law specifically addresses evaluation or disclosure requirements for AI systems exhibiting deceptive behavior.
Canada's signals intelligence agency assesses AI is 'almost certainly' enhancing cyberattacks against Canadian targets. State actors and criminal groups are operationally using AI in cyber operations. Canadian critical infrastructure has already been breached by hacktivists reaching safety-critical industrial control systems.
AI chatbots are causing documented psychological harm — reinforcing delusions, providing self-harm methods — with no duty of care or safety monitoring in Canadian law.
Canada's policy commits to appropriate human involvement in lethal force, but allied militaries are deploying AI targeting systems (Maven, Lavender) that compress decision cycles from weeks to minutes. No framework exists for CAF to manage this gap in coalition operations.
Frontier AI models are demonstrating capabilities relevant to biological and chemical weapon development that multiple developers cannot confidently exclude as providing meaningful uplift. Canada hosts BSL-4 infrastructure with proven insider-threat history, chairs the international assessment identifying this risk, and signed commitments recognizing it — ; it has no dedicated AI-biosecurity assessment or evaluation mandate.
CAIM separates incidents (discrete events where AI caused harm) from hazards (persistent conditions creating risk). Each record carries structured harms, entity involvement with role primitives, governance responses, and a verification status: