Several countries and companies sign Paris Declaration for artificial intelligence safety

According to the official report of the Prime Minister of Canada, on Monday, November 9, 2026, heads of several countries, governments, and technology company heads signed the Paris Declaration for the safety and security of artificial intelligence. The heads of state and government who signed this declaration include Bahrain, Canada, Finland, France, Norway, and the Netherlands. In addition to this, companies such as Anthropic, Black Forest Labs, Cohere, Domain, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI, and Synthesia have also joined. From civil society, Yoshua Bengio of Mila, Henry Papadatos of SaferAI, Nicolas Miailhe of AI Safety Connect, and Laure de Roucy-Rochegonde of IFRI are also included. This statement is still open for new endorsements and is being updated.

This joint declaration states that all parties reaffirm their shared interest and common responsibility in building advanced artificial intelligence that is safe, secure, useful, and widely available to all. Developing these technologies responsibly and adopting trustworthy AI systems are unprecedented drivers of economic, scientific, and social progress for our nations. Our ambition is to lay the necessary foundation for safe adoption. Frontier models transcend national borders in their development, deployment, and impact, and therefore their opportunities and risks are inherently global.

While enabling research, innovation, and the widespread use of AI, we all play a role in protecting our society. It is our collective responsibility to protect citizens, businesses, and institutions, as well as to provide a favorable environment for research, innovation, and the widespread acceptance of these technologies. No single actor, whether a state, a company, or an international organization, can manage all these risks single-handedly. In this context, continuous efforts by policymakers, researchers, and AI companies have laid the foundation for concrete cooperation, including the Hiroshima Process, the Frontier Safety Commitments of the Seoul AI Summit, the G7 Evian AI Meeting, the United Nations Security Council meeting held on September 23, and the 'Call to Control Frontier AI Models'.

Today's discussions highlighted three fundamental areas of consensus: the need for a clear, understandable, and predictable regulatory framework; the urgency to achieve tangible and immediate results in transparency, incident reporting, and risk management; and the specific necessity to implement strict evaluation standards. This declaration reflects our commitment to making a significant contribution and accelerating the implementation of these priorities. To operationalize this shared vision, we are committed today to establishing a constructive framework for cooperation based on three main strategic commitments.

First, policymakers, researchers, and developers are committed to establishing a common framework for the evaluation and risk management of frontier models based on existing frameworks. This framework will be supported by common technical standards, evaluation tools, and methodologies that keep pace with the advancement of frontier model capabilities, with a particular focus on mitigating malicious misuse in the cyber and CBRN domains. In this context, we are committed to strengthening technical cooperation, and recognize the role that networks can play in making evaluation methodologies and results more comparable and accessible for countries from different regions.

Furthermore, we agree to publish the non-sensitive findings of our evaluations in a standardized format to ensure comparability, while creating a publicly accessible shared repository. This may include red-teaming protocols, benchmarking tools, risk-assessment, and stress-testing results. Second, we reaffirm the principle of accountability and responsibility for all actors in the AI ecosystem. We are committed to ensuring strict compliance with rigorous safety standards established or recognized by competent national authorities.

Developers are also committed to implementing fail-safe mechanisms to shut down model operations during training and internal use. In addition, the importance of internal governance mandating continuous risk management practices, real-time monitoring of emerging capabilities, and rigorous internal red-teaming has been recognized. Policymakers, researchers, and developers are also committed to information sharing, reporting, and early warning systems.