France publishes a Paris Call on frontier AI safety signed by twelve governments and ten companies

The Élysée published an English text of the Paris Call for AI safety and security, posted on 9 November 2026. The page says the declaration was signed on Monday 9 November 2026. It lists these heads of state and government: Bahrain, Canada, Finland, France, Norway, the Netherlands, Cyprus, Kenya, Latvia, Austria, Estonia and Ireland. It lists these companies: Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia. It also lists civil-society signatories: LawZero, SaferAI, AI Safety Connect, Ifri, Geopolitical Center for the Datasphere, GPAI Policy Lab and CeSIA. It says the statement remains open to new endorsements and is being updated.

The page sets out three commitments. The first is a common framework for evaluating and managing risks from frontier models, including dangerous misuse in the cyber and CBRN domains and risks from growing autonomy or misalignment, including loss of control, and publication of non-sensitive evaluation findings in a standardised format. The second is responsibility according to each actor's role. It says developers of frontier models commit to disclose model capabilities, limitations and known risks and, where appropriate, to support evaluation by qualified independent evaluators. It says they also commit to implement fail-safe mechanisms to shut down model operations within their control, including during training and internal use, and it gives restrictions, pauses or shutdowns as examples. The third is information sharing, reporting and early warning. It says policymakers commit to offer secure and confidential channels for whistleblowers to raise concerns about systemic risks without fear of legal or professional retaliation. The page does not set a penalty or a date by which a duty starts.

What it means for business: this is a signed declaration, not a statute. The company commitments on disclosure, independent evaluation and shutdown mechanisms are written as commitments by the developers named on the page. The page does not create a filing, a fine, or a deadline. This describes the declaration. It is not advice on a model.