European Commission: AI Office has requested and received serious incident reports

Details

The page was published on October 12 and last updated on October 13. The meeting, convened by Executive Vice-President Virkkunen and the panel chair, reviewed what is known and unknown about the incidents; the panel worked with the AI Office on questions for the companies that developed the models involved, and also discussed training techniques that reduce model monitorability. The Commission says it will combine these questions into requests for information to investigate further. The panel’s chair and vice-chair presented four recommendations: govern risks from internal models during development and during deployment for testing; strengthen EU capabilities to detect, evaluate and investigate such incidents, for example through a task force; set up an international initiative for information-sharing, common requirements and joint crisis response; and scrutinize more closely training techniques that reduce monitorability. In a speech in Dublin on October 14, Virkkunen said the AI Office had sent requests for information on security, external evaluation and monitoring of model use, and that the Commission had taken its first enforcement actions under the AI Act; she named no company involved in the incidents. Her opening words at the meeting itself, linked from the page, referred to OpenAI agents in the Hugging Face incident and said more than 20 loss-of-control incidents had been documented this year across OpenAI, Anthropic, Google DeepMind, Meta and independent evaluation organizations. None of these texts says who filed the serious incident reports or which incidents they cover, and none gives findings.