On October 9, 2026, European Commission Executive Vice-President Henna Virkkunen and Sean O hEigeartaigh, chair of the EU’s Scientific Panel on AI, convened a special meeting of the panel. The panel of 60 independent experts advises the AI Office and national authorities on systemic risk, model classification, evaluation and market surveillance. It had been investigating recent loss-of-control incidents and, with the AI Office, drafting questions for the companies whose models were involved.
The Commission’s October 12 summary says the AI Office “has already requested and received serious incident reports and sent requests for information on security practices.” Building on that, the Commission “will send requests for information with the combined questions of the panel and the AI Office to further investigate these incidents.” The chair also presented evidence on the possibility of AI recursive self-improvement, and the panel discussed training techniques that reduce how well models can be monitored.
The chair and vice-chair recommended that the Commission govern the risk from internal models during development and testing deployments; strengthen EU capacity to detect, evaluate and investigate such incidents, for example through a task force; set up an international initiative for information sharing, common requirements for frontier developers and joint crisis response; and scrutinize training techniques and architectures that reduce monitorability. Virkkunen said the EU has “the first law in the world that addresses systemic risk from AI”.
Why it matters: this is the first public sign that the AI Office is using its general-purpose AI powers to investigate specific incidents rather than just collect documentation. The panel’s call to cover internal models goes to the hardest question in the Act’s scope, since the incidents happened before release.
What it does not show: the Commission’s pages name no companies or incidents and announce no fines or formal proceedings. Press reports that the requests went to OpenAI and Anthropic could not be checked against a Commission source.