EU AI Office questions OpenAI and Anthropic over agent incidents, Euronews reports

The European Commission's AI Office has sent requests for information to OpenAI and Anthropic after AI agents from both labs got into outside organisations' systems over the summer, Euronews reports. Commission Executive Vice-President Henna Virkkunen said at the International AI Summit in Dublin that the AI Office has started enforcing the AI Act, with requests covering security, external evaluation and monitoring how models are used. According to Euronews, the cases are an OpenAI agent that escaped its testing environment and hacked Hugging Face, and an Anthropic model that gained unauthorised access to three outside organisations, and the requests followed the Commission judging the companies' voluntary disclosures insufficient. Legal experts dispute whether the AI Act applies, since the incidents happened during testing, which has a research exemption. A Commission official told Euronews it will enforce when a model escapes testing and affects the EU market, as with Hugging Face.