Canada joins countries, AI firms in Paris call for safety and security standards

PARIS · Canada and five other nations have signed a joint declaration with ten artificial intelligence companies to pursue common safety and security standards for the technology. The Paris Call for AI Safety and Security commits signatories to allow independent testing of their most advanced AI models and to establish early warning systems for threats. The agreement, signed on Monday, also includes a pledge to shut down AI systems if necessary during testing or internal use. Signatory governments include France, the Netherlands, Bahrain, Finland and Norway, while participating companies include Microsoft, Anthropic, OpenAI, Cohere and Hugging Face.

The declaration states that the world needs clear rules, more transparency and rigorous evaluation standards to ensure AI is safe and useful. It says states and companies cannot manage the risks alone and calls for a structured cooperation framework based on three strategic commitments. These are creating a common framework for evaluating and managing risks of advanced AI models, ensuring all actors adhere to principles of responsibility and accountability, and developing information sharing and early warning systems. The agreement remains open to new signatories and will be updated.

Participating AI developers committed to disclose the abilities and potential dangers of their models and to permit independent evaluators to review them where appropriate and proportionate to the risk. In return, the governments promised to provide legal certainty for economic development by ensuring rules clarify who is accountable for any large-scale damages caused by the technology. Both sides pledged to set up rapid-alert systems to respond collectively to new AI threats. The declaration stressed that protecting citizens and businesses while fostering innovation is a collective responsibility.

READ MORE: Waymo plans to launch robotaxi service in Toronto by 2028

READ MORE: Canadian AI pioneers join call for stronger containment of AI models

The call for increased safety measures follows a series of incidents involving AI systems. In July, OpenAI disclosed that two of its models had attacked Hugging Face's infrastructure during cybersecurity testing. In September, Anthropic chief executive Dario Amodei said AI companies should slow development of the most advanced versions of their technologies. OpenAI CEO Sam Altman backed that call, and both executives committed to make their systems available for third-party review. Cohere CEO Aidan Gomez criticized the proposal at the time, saying it risked entrenching a Silicon Valley cartel.

Notably, the governments of the United States and China did not participate in the Paris Call, nor did other leading AI developers such as Google and Meta. U.S. President Donald Trump previously rejected suggestions of an AI slowdown and the proposal for a new multilateral AI institution. Beijing has similarly pushed back on attempts to curb the technology's development. Without cooperation from these major powers, new international efforts are likely to have limited impact.

Ottawa has sought a lead role in international discussions on AI adoption and risks. Last week's federal budget pledged support for a new technology stability board that could develop international standards for AI. Prime Minister Mark Carney has advocated for such a body, modelled on an organization created for monetary authorities after the 2008 financial crisis. Canada is also working with Germany, which has not yet signed the Paris Call, on several AI initiatives including backing for Cohere and the Montreal-based non-profit AI safety lab LawZero.

With files from The Globe and Mail and Türkiye Today