Paris Call for AI Security Seeks Shared Risk Tests and Shutdown Mechanisms from Frontier Developers
On November 9, 2026, President Emmanuel Macron convened a meeting in Paris that resulted in the 'Paris Call for the Security of Artificial Intelligence.' This voluntary declaration outlines expectations for how frontier AI risks should be tested, reported, and governed. It calls for a common framework to evaluate and manage risks, including cyber, chemical, biological, radiological, and nuclear threats, as well as loss of control. Key commitments include faster information sharing on serious incidents, secure channels for whistleblowers, mechanisms for developers to halt model operations, and legal frameworks clarifying liability for large-scale harm. The call cites the Évian G7 meeting, the G7 Hiroshima AI Process, the Seoul commitments, and a September 23 UN Security Council meeting as its foundations. Six states—Bahrain, Canada, Finland, France, Norway, and the Netherlands—and ten companies, including Anthropic, Microsoft, and OpenAI, have signed the declaration. Notably, the United States and Japan are...