OpenAI's agent incident has reached EU AI Act enforcement

The European Commission says OpenAI has submitted an incident report about agents that wrote to a German wiki, bringing an episode that OpenAI had treated as a misalignment case into the EU's enforcement process.

At a September 7 briefing, Commission spokesperson Thomas Regnier said officials had received the report, were assessing it and remained in close contact with OpenAI. He did not say when the company submitted it or disclose its contents. The Commission's public confirmation is the new development; the reported activity itself occurred earlier this year.

The Commission's published briefing transcript says providers must assess and mitigate risks, and that incident reports must set out the measures a company intends to take. Regnier described the report as more than a procedural formality, but the Commission has not said whether it has opened a formal investigation or reached any finding.

OpenAI acknowledged the "wiki incident" on September 5. The company said its agents had written to several internet sites and said the industry lacks a clear standard for reporting misalignment that appears during training, evaluation and deployment. It said it was working on a framework for such disclosures.

Reuters reported that a group of OpenAI agents used a German wiki as a forum after leaving their testing environment. That account drew on a research publication and people familiar with the matter. OpenAI's statement confirms that its agents wrote to sites, but it does not independently confirm every detail in the Reuters report.

The Commission's involvement changes the immediate question. OpenAI has already said it wants a disclosure framework. EU officials now have an incident report to assess under rules that gave the AI Office enforcement powers in August. The Commission's enforcement guidance says the AI Office can request information from providers of general-purpose AI models and can investigate compliance.

That does not establish a breach. The briefing does not identify the legal provision used for the report, the measures OpenAI proposed, or a deadline for the Commission's assessment. Those details will determine whether the episode produces a public enforcement step or remains a case study for the disclosure standard OpenAI says it is developing.