EU Will Query AI Labs Over Recent Loss-of-Control Incidents

The European Commission will send requests for information to further investigate recent loss-of-control incidents involving AI models, it said in a news article published on 12 October 2026, after a special meeting of the AI Scientific Panel on 9 October.

The Commission says the panel has been investigating the incidents and, with the AI Office, worked on a set of questions for the companies that developed the models involved. The panel’s chair also presented four recommendations on frontier AI safety and security.

What the Commission Says Happened at the Meeting

According to the Commission, Executive Vice-President Henna Virkkunen and Seán Ó hÉigeartaigh, Chair of the AI Scientific Panel, convened the special meeting on 9 October 2026. It aimed to inform the Commission’s actions in view of recent incidents of loss of control, and technical developments related to monitorability of models and recursive self-improvement.

The Commission says the chair and vice-chair presented what is known about the incidents to date and highlighted what is still not known. It says the AI Office has already requested and received serious incident reports and sent requests for information on security practices.

The chair also presented evidence related to the possibility of AI recursive self-improvement and related risks, the Commission says, and the panel discussed recent developments in training techniques which reduce the monitorability of models.

The Four Recommendations

The Commission lists the chair’s and vice-chair’s recommendations to the Executive Vice-President and the AI Office. The first is to ensure governance of the risk posed by internal models during development activities and deployment for testing.

The second is to strengthen EU capabilities related to detection, evaluation and investigation of incidents of this nature, for example via a task force, preparing the technical and procedural groundwork for future investigations.

The third is to establish an international initiative to facilitate information-sharing and common requirements for frontier AI developers, as well as joint crisis response and prevention. The fourth is to ensure enhanced scrutiny of training techniques and architectures that reduce monitorability, given the risks they present.

Who Sits on the Panel

The Commission’s 9 October press release says the Scientific Panel brings together 60 independent experts who advise the EU AI Office and national authorities on systemic risks, model classification, evaluation methodologies, and cross-border market surveillance.

Virkkunen said in that release: “The EU has the first law in the world that addresses systemic risk from AI, and we need state-of-the-art scientific input.” The Commission’s follow-up came the same day the UK government set out new steps on AI incidents and agent risks.

The Structural Read

As we read it, the first recommendation reaches inside the labs: the chair asks for governance of the risk posed by internal models during development and deployment for testing, not only models on the market.

As we read it, monitorability is now a policy term: the Commission says the panel discussed training techniques which reduce the monitorability of models, and the chair recommends enhanced scrutiny of them.

As we read it, the incidents themselves remain undisclosed: the Commission’s texts describe questions for the companies that developed the models involved, without naming them.

Three Implications

FRONTIER AI DEVELOPERS The Commission says it will send requests for information combining the panel’s and the AI Office’s questions to the companies that developed the models involved.

INTERNAL AND TEST MODELS The chair recommends governance of the risk posed by internal models during development activities and deployment for testing.

INTERNATIONAL COORDINATION The chair recommends an international initiative for information-sharing, common requirements for frontier AI developers, and joint crisis response and prevention.

The Business Engineer Lens

This story maps onto the Business Engineer framework Inside Anthropic’s Permission Layer.

The framework puts it this way: “The governance layer sits at the intersection of capability, regulation, national security, access control, and geopolitical alignment.”

As we read it, the panel’s four recommendations touch several of those at once: internal-model governance on capability, incident investigation on regulation, and an international initiative on alignment between governments.

What Is Not Established

The two Commission texts we read do not name the companies, the models or the incidents, and do not describe what happened in them. They do not say when the new requests for information will be sent, or whether the Commission will adopt the chair’s recommendations.

The Bottom Line

The Commission says it will send requests for information, combining questions from the AI Scientific Panel and the AI Office, to further investigate recent loss-of-control incidents. The panel’s chair recommended governance of internal models, stronger EU incident investigation, an international initiative and scrutiny of less monitorable training techniques.