Council of Europe celebrates AI convention milestone with Zurich summit

The Council of Europe’s AI Framework Convention is not as flashy as a new model release, but it is one of the few concrete attempts to bind AI development to human-rights law across borders. Two years in, the treaty has 21 signatories, including the EU, US, Canada, Japan, Switzerland and Uruguay, and the Union’s ratification in May 2026 turned it into a live legal instrument rather than a symbolic gesture. The Zurich conference the Council is convening with ETH’s Albert Einstein School is about the unglamorous follow-through: how to translate high-level principles into procurement rules, liability standards and oversight bodies.

For the race to AGI, the convention matters because it is a template for how liberal democracies might try to steer frontier systems without shutting them down. The text is risk- and context-based rather than outright prohibitive, which gives governments room to demand tighter scrutiny of Astra-class systems and agentic deployments while leaving room for lower-risk applications. The competitive question is whether signatories can move fast enough on implementation that their firms are not structurally disadvantaged against jurisdictions that prioritize speed over safeguards. If the Council’s process yields credible, interoperable standards, it could actually lower legal uncertainty for large players, making it easier to scale advanced models into regulated sectors like health, finance and public services.