EU science panel meets on AI loss-of-control incidents and monitoring

In short

The European Commission says its AI Scientific Panel held a special meeting on 9 October 2026 about recent "loss-of-control" incidents and related technical developments. The Commission's report does not name the models or describe what happened in the incidents. The Commission says it will send companies further questions as part of the investigation.

What happened

The Commission published the original report on 12 October 2026. It says the meeting was convened by Executive Vice-President Henna Virkkunen and the panel's chair, Seán Ó hÉigeartaigh. Its aim was to inform the Commission's actions.

According to the Commission, the panel has been investigating the recent incidents. At the meeting, the chair and vice-chair presented what is known so far and highlighted what is still unknown.

The report does not define "loss of control" or give details of any incident. It does not say whether the incidents happened during training, testing, simulation or real-world use. It also does not say whether any person or organisation was affected.

The panel also discussed two technical topics:

• Recursive self-improvement. This generally means AI systems helping to improve AI systems, including themselves. The chair presented evidence about whether this is possible and about related risks.
• Monitorability. This means how easily people can observe and check what a model is doing. The panel discussed recent training techniques that make models less monitorable, and what this could mean for future incidents and efforts to reduce risk.

What was done about it

The Commission says the AI Office has already requested and received serious incident reports. It has also asked for information on security practices. The panel and the AI Office have drawn up a set of questions for the companies that developed the models involved, and the Commission says it will send these as formal requests for information.

The chair and vice-chair also made four main recommendations:

• Ensure governance of the risk from internal models during development and while they are deployed for testing.
• Strengthen EU abilities to detect, evaluate and investigate such incidents, for example through a task force.
• Set up an international initiative for sharing information, common requirements for developers of frontier AI (the most advanced systems), and joint crisis response and prevention.
• Look more closely at training techniques and model designs that reduce monitorability.

Why it matters

The report says the meeting was meant to inform the Commission's actions in light of these incidents and technical developments. The panel's leaders recommend closer scrutiny of techniques that reduce monitorability, "given the risks they present". The investigation is still under way, and the report says there are still things the panel does not know about the incidents.