aipopdb

22 Nov 2026  ·  10,137 works · 1,192 news indexed

News / Seventeen governments and ten AI developers sign the Paris Call on frontier AI safety

Seventeen governments and ten AI developers sign the Paris Call on frontier AI safety

AIPopDB Newsroom · via elysee.fr · · 2 min read · Comment

Read the original ↗
Image for Seventeen governments and ten AI developers sign the Paris Call on frontier AI safety

Image: Élysée

A declaration called the Paris Call for AI safety and security was signed on Monday 9 November 2026 by heads of state or government, AI company chief executives and civil-society groups. Its text says the day's talks produced agreement on three points: a clear and predictable regulatory framework, concrete near-term results on transparency, incident reporting, risk management and verification, and rigorous standards for evaluating models.

“Frontier models transcend national borders in their development, deployment and impact, and their opportunities and risks are therefore inherently global.”

The Paris Call declaration text, in the declaration text · elysee.fr ↗

Seventeen governments signed: Austria, Bahrain, Canada, Croatia, Cyprus, Estonia, Finland, France, Germany, Ireland, Kenya, Latvia, Liechtenstein, Luxembourg, the Netherlands, Norway and Portugal. Ten companies joined them: Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia. Ten civil-society signatories include LawZero, SaferAI, CeSIA, FAR.AI and The Future Society, along with AI Safety Connect, Ifri, the GPAI Policy Lab and the Oxford Martin AI Governance Initiative.

The first commitment is a common framework for evaluating and managing the risks of frontier models, with shared technical standards. It centres on misuse, such as cyber and chemical, biological, radiological and nuclear threats, and on loss of control. Non-sensitive evaluation findings would be published in a standard format in a public shared repository. The text names the NAAIMES evaluation network as one route to making results comparable across regions.

The second commitment is accountability across the AI ecosystem, through compliance with safety standards that national authorities recognise and independent third-party evaluation. Developers promise to disclose what their models can do, where they fall short and what risks are known, and to build fail-safe shutdown mechanisms that work during training and internal use. They also acknowledge that safeguards can be stripped from open-weight models, while saying the research benefits of open weights should be kept. Governments say legal frameworks should make clear who is responsible for large-scale damage, including across borders, so that AI firms have legal certainty.

The third commitment covers information sharing and rapid alerts for serious cyber incidents and incidents involving loss of AI control, with governments offering confidential channels for whistleblowers. The signatories will explore whether a cooperation mechanism for sharing incident information is needed and what it should cover. The text builds on the Hiroshima process, the Seoul Summit Frontier Safety Commitments, the G7 meeting in Evian and the UN Security Council's AI meeting of 23 September. The statement remains open to new endorsements and is being updated.

Sources · 1

Comments