Ten AI Firms Sign Paris Call on AI Safety, Pledging Shutdown Mechanisms Without Penalties

What happened

The declaration was initiated by French President Emmanuel Macron and signed the day before the Paris Peace Forum opened; about 50 people joined the discussion by video, including OpenAI President Sam Altman, Anthropic President Dario Amodei, Microsoft President Brad Smith and U.S. Treasury Secretary Scott Bessent. The first governments to sign were Bahrain, Canada, Finland, France, Norway and the Netherlands; the ten companies were Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia, joined by civil-society groups including Yoshua Bengio's LawZero.

The declaration centers on three areas of cooperation: common approaches to evaluating and managing frontier model risks, clearer responsibility and accountability across the AI ecosystem, and stronger information sharing and early warning. Signatory companies pledge to disclose known risks, support risk-proportionate independent audits, report serious security incidents faster, and set up fail-safe mechanisms during training and internal use, including restrictions, pauses or shutdowns of the models they control.

The declaration creates no binding legal obligation and sets no deadlines or sanctions; the U.S. and Chinese governments did not sign, nor did Google or Meta. French AI safety organization CeSIA signed the declaration while arguing that it leaves crucial gaps. The declaration remains open to new endorsements, with the list being updated.

Key facts

1. Signed: Nov 9, 2026, the day before the Paris Peace Forum opened; initiated by French President Emmanuel Macron
2. Governments: First six: Bahrain, Canada, Finland, France, Norway and the Netherlands; the declaration stays open and the list is updated
3. Companies: Ten: Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia
4. Three pillars: Common evaluation methods for frontier model risks, clearer responsibility, and information sharing with early warning
5. Company pledges: Disclose known risks, support independent audits, report serious incidents faster, and set up fail-safe mechanisms including restrictions, pauses or shutdowns
6. Not signed: The U.S. and Chinese governments, Google and Meta did not sign; the declaration has no deadlines or penalties