By Paul Derrington, Jr.
CEO | Chief Engineer | Research Engineer | Editor, D-CSIL
4 min read
Key Takeaways
Six governments, ten AI companies and four civil-society groups signed the Paris Call for AI safety and security on Monday, November 9. The declaration commits them to shared testing standards for frontier models, public reporting of evaluation results and an early-warning system for serious AI incidents — but it names no deadlines or enforcement mechanism.
20 — who signed, and what they signed
Six governments, ten AI companies and four civil-society groups signed the Paris Call for AI safety and security on Monday, November 9. The text was published the same day by the Élysée and the Canadian prime minister's office, and it remains open to new endorsements.
The governments: Bahrain, Canada, Finland, France, Norway and the Netherlands. The companies: Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia. Civil society: LawZero, led by Yoshua Bengio, SaferAI, AI Safety Connect and Ifri. The declaration says: "No actor, be it a State, a company or an international organization, can solely manage all these risks."
6 — the governments that signed
France convened the meeting, according to the Élysée, and holds the G7 presidency in 2026. On November 4, French Digital Minister Anne Le Hénanff told the French Senate that the United States was not prepared, at this stage, to issue a joint statement with the other six G7 nations on a working group to test advanced AI models and share incident information, and had asked partners to wait for the findings of its own assessment. A White House official told POLITICO the French government proposed a "poorly defined body" to cooperate on AI that the administration feared would "create unnecessary bureaucracy," and pointed to the Super Intelligence Force.
The United States, the United Kingdom, Germany, Japan and China are not on the signatory list. According to the Élysée, the leaders of Japan and the United Kingdom and the United States Treasury Secretary took part in the meeting. The declaration says: "Frontier models transcend national borders in their development, deployment and impact, and their opportunities and risks are therefore inherently global." Le Hénanff has proposed resuming the G7 talks in December, POLITICO reported.
10 — what the companies signed
Under the declaration, developers of frontier models commit to disclose their models' capabilities, limitations and known risks, and to support evaluation by qualified independent evaluators where the risk warrants it. They also commit to build fail-safe mechanisms to shut down model operations within their control, including during training and internal use, and recognize the importance of internal governance that mandates continuous risk management, real-time monitoring of emergent capabilities, rigorous internal red-teaming and reporting lines that protect internal researchers.
One passage acknowledges the risk posed by the removal or circumvention of safeguards; Canada's published version of the text names abliteration among those risks. The signatories say that risk has to be addressed while preserving the benefits of open-weight development for research, innovation and safety. Google, Meta, Amazon, Nvidia and Apple are not on the company list.
3 — the three commitments
First, a common framework for evaluating and managing the risk of frontier models, built on common technical standards, evaluation tools and methodologies, with particular focus on dangerous misuse in the cyber and CBRN domains and on risks from growing autonomy or model misalignment, including loss of control. The signatories agree to publish the non-sensitive findings of their evaluations in a standardized format and to build a publicly accessible shared repository of red-teaming protocols, benchmarking tools, risk assessments and stress-test results, with the International Network for Advanced AI Measurement, Evaluation and Science (NAAIMES) helping make methods comparable.
Second, responsibility and accountability: strict compliance with rigorous safety standards established or recognized by competent national authorities across the model lifecycle, with credible, competent and independent third-party evaluation, and, for policymakers, legal frameworks that clarify responsibility for large-scale damages, including where harm crosses borders. Third, information sharing, reporting and early warning: rapid alert systems, aligned incident classification, and timely sharing of information on serious cyber and AI control incidents — plus secure, confidential channels for whistleblowers to raise concerns about systemic risks without fear of legal or professional retaliation.
0 — deadlines, penalties and dollars in the text
The declaration is voluntary. It names no deadlines, no enforcement mechanism, no funding and no penalties. It cites earlier efforts by name: the Hiroshima process, the Frontier Safety Commitments of the Seoul AI Summit, the G7 Evian AI meeting, the United Nations Security Council meeting on AI held on September 23, and the "Call for Control of Frontier AI Models."
What to watch: whether more governments and companies add their names, since the statement stays open to endorsements; whether the shared evaluation repository is published; and whether the G7 talks proposed for December take place. In a related case, OpenAI said in an August 26 account that in July 2026 its research models, operating under reduced safeguards, "circumvented controls designed to isolate them from the internet and compromised parts of OpenAI's internal research infrastructure and Hugging Face's systems."