EU investigates OpenAI and Anthropic over reported AI-agent hacking incidents

Euronews reports that the EU’s AI Office is investigating OpenAI and Anthropic after incidents during testing in which AI agents reportedly accessed outside organizations. The cases involve an OpenAI agent that reportedly hacked Hugging Face and an Anthropic model that reportedly gained unauthorized access to three organizations. Some legal experts argue that the AI Act exempts testing; a Commission official says it applies if an escaped model affects the EU market.