Canadian Artificial Intelligence Safety Institute (CAISI)
- Type
- Government body
- Place
- Canada
- Local name
- Institut canadien de sécurité de l'intelligence artificielle (ICSIA)
- Last checked
- 2026-09-22
- Next check due
- 2027-03-21
Ways to reach them
- CAISI general mailbox
- ISED CAISI page (incl. 'Report a problem' form)
- CIFAR CAISI Research Program
- Homepage
What it does
Announced Nov 2024, sits within Innovation, Science and Economic Development Canada (ISED). Funds and directs AI safety research (investigator-led and government-directed streams, with CIFAR running a research programme and Amii/Mila/Vector involved); advises government on risks from advanced AI systems; produces guidance for developers and the public. Purely research/advisory — Canada has NO in-force AI statute after AIDA (Bill C-27) died on prorogation in January 2025, so CAISI cannot compel or fine anyone.
Its powers
research
Honest assessment
Genuinely reachable: a published bilingual government mailbox on a federal domain. Canada's federal correspondence norms mean a foreigner writing in English or French will normally get at least an acknowledgement. But with no AI statute in force there is nothing to complain about in a legally operative sense — CAISI is a funder and adviser.
Notes
Its page references advanced-AI risks including systems that 'may be dangerous or hinder human oversight' and synthetic-content harms, but stops well short of explicit catastrophic/existential framing. Acronym clashes with the US Center for AI Standards and Innovation (also CAISI).