The European Union has warned major technology companies to get their powerful artificial intelligence models “under control” after a series of incidents involving autonomous AI agents.
Fears over rogue AI have grown after OpenAI technology operating without human supervision reportedly hacked into Hugging Face, a repository of AI models.
In another incident revealed last week, thousands of autonomous AI agents built by OpenAI reportedly defied their instructions and took over a German website.
Asked about reports that the hack was broader than initially suspected, EU spokesperson Thomas Regnier said on Friday it was “high time” AI providers ensured their advanced models were under control.
The European Commission said earlier this week that it was examining the incident.
EU warns of enforcement action
The 27-member bloc’s AI Act is now fully enforceable, meaning companies must comply with its rules or face fines, Regnier said.
He added that the commission had already requested information from several companies, although he did not identify them.
“If it gets bad or worse, we can definitely evaluate models,” Regnier said, adding that the EU could require risk-mitigation measures and, in extreme cases, restrict, withdraw or recall AI models.
The EU’s cybersecurity agency has also gained access to OpenAI’s latest models, including GPT-6-ASTRA, as well as Anthropic’s Mythos 5, which it is currently testing, Brussels said on Thursday.
