The Race — Building Democracy in the Age of AI. People race toward a shared finish line.
The Race
← Trends

Supporting case

Governments, companies and civil society sign the Paris Call on AI safety

What happened

On November 9, signatories to the Paris Call pledged to share information about serious AI incidents and risks and cooperate on safety standards and testing. Canada’s published signatory list names Bahrain, Canada, Finland, France, Norway and the Netherlands, alongside model developers and civil-society organizations. The meeting had broader attendance than that list; attendance alone does not show a government signed. The joint declaration ↗ says developers should disclose known capabilities and risks, support qualified independent evaluations when proportionate to the risk, and maintain ways to stop model operations under their control. It proposes technical cooperation through existing networks, including the International Network for Advanced AI Measurement, Evaluation and Science. The declaration does not set enforceable reporting deadlines, designate who can compel access, or specify penalties.

Shared warnings and outside testing could reduce how much each government depends on individual companies for information. The declaration leaves considerable discretion to developers and participants. The practical test is whether a smaller regulator or independent evaluator can obtain useful information in time to act.

Dates and implementation stage

November 9, 2026: Paris Call issued.

Supporting sources

Related trends