Initiative overview
Announced in November 2024, the Canadian AI Safety Institute (CAISI Canada) is Canada's domestic counterpart to equivalent bodies established in the UK, the United States, Japan, and other countries participating in the Hiroshima AI Process. Operating within Innovation, Science and Economic Development Canada (ISED), it focuses on three core activities: technical evaluation of advanced AI models to identify and characterise safety risks; development of safety testing methodologies and benchmarks; and international coordination through the International Network of AI Safety Institutes (INAIS). CAISI Canada contributes Canadian expertise and perspectives to the global AI safety research agenda, including work on frontier model evaluations, systemic risk assessment, and the development of safety standards. It plays a bridging role between the Canadian government, AI research institutions (including Mila, the Vector Institute, and Amii), and international partners, ensuring that Canadian AI safety research informs both domestic AI governance frameworks and international norms. The Institute was established in direct response to commitments made by Canada at the Bletchley Park AI Safety Summit in November 2023, and formalises Canada's participation in the international AI safety evaluation network.
