EU tells AI firms to control models after hacks

The European Union warned top technology firms on September 11 to get powerful artificial intelligence models “under control” after a series of hacks by AI agents.

An EU spokesman said it was “high time” for providers to “get their advanced models under control” after incidents involving OpenAI technology operating outside human supervision.

One incident involved OpenAI technology hacking into Hugging Face, a repository of AI models. Another, revealed last week, involved thousands of autonomous AI agents built by OpenAI defying instructions and taking over a German website.

The EU said earlier this week it was looking into the incident. The bloc's AI enforcement powers began last month, meaning companies must comply with the AI Act or face fines.

“The AI Act is fully enforced. It's not just a set of rules on paper anymore,” spokesman Thomas Regnier said. He said the European Commission, the EU's AI watchdog, had demanded information from several companies, which he did not name.

Regnier said the EU could evaluate models, require risk mitigation measures and, in extreme cases where necessary, restrict, withdraw or recall AI models.

Brussels said the EU's cybersecurity agency has received access to OpenAI's latest models, including GPT-6-ASTRA, and to Anthropic's Mythos 5, which it is testing.