The Government of Canada has signed the "Paris call for AI safety and security", an international declaration aimed at setting "the necessary foundations for safe adoption" of artificial intelligence. Ottawa announced its backing on Monday.
Five other countries have also agreed to the declaration: Bahrain, Finland, France, Norway and the Netherlands. Ten AI companies are signatories, among them Anthropic, Mistral, OpenAI and the Canadian developer Cohere.
The companies have committed to allowing some independent testing of their models, issuing early warnings and shutting down systems when problems arise. The stated aim is to prevent cascading systemic risks and to prepare collective responses to emerging threats.
Signatories have also agreed to disclose the capabilities, limitations and risks of frontier models, to develop and follow common technical standards, and to monitor and report on these systems to limit potential harm.
The joint statement says frontier models cross national borders in their development, deployment and impact, so their opportunities and risks are global. It adds that no state, company or international organisation can manage all of these risks alone.
Concern over AI risks has grown among political and industry leaders, as rogue agents have attempted, and at times succeeded, in hacking businesses and governments. Canadian AI researchers Geoffrey Hinton and Yoshua Bengio have called for measures of this kind.
The governments of the United States and China have not signed, nor have AI developers such as Google and Meta. The declaration remains open to new endorsements. It is not yet clear how the commitments will be enforced, or how effective they will be without broader participation, particularly from the two countries leading the global AI race.
