Regulatory  ·  2026-11-20

European Commission GCSA and EGE issue unprecedented joint statement urging sovereign AI model-evaluation capacity and resisting AI Act dilution

RegulatoryMedium impactEuropean Union
On 16 November 2026, the European Commission's Group of Chief Scientific Advisors and the European Group on Ethics in Science and New Technologies issued a joint statement, 'The next phase of artificial intelligence — Governing for safety, security and human agency'. It warns traditional public-policy guardrails cannot govern rapidly advancing AI and sets four recommendations: keep existing rules strong (rigorous enforcement of the AI Act, Regulation (EU) 2024/1689, and resisting lobbying to dilute compliance deadlines); calibrate model development/deployment to degree of danger, contingent on verified mitigation for severe cyber/biosecurity/critical-infrastructure risks; build independent public evaluation capacity (sovereign testing facilities rather than commercial self-attestations); and establish an ecosystem for continuous learning (adaptive monitoring that ingests operational deployment evidence).
Top EU scientific and ethics advisers are signalling the AI Act must be enforced rather than diluted, and that pre-deployment safety cases and sovereign public model-evaluation capacity are needed. This is a strong institutional signal against regulatory backsliding and in favour of independent public benchmarking/testbeds, which would materially shape future EU implementing rules, the general-purpose AI Code of Practice and the AI Act's evaluation/transparency obligations.
EU AI developers, deployers and AI-purchasing institutions should anticipate tighter enforcement and possible sovereign evaluation mandates; track the general-purpose AI Code of Practice and any Commission follow-up to the joint statement.
CASRAI report on the joint statement
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →