September 11, 2026, 15:34 GMT | Insight
The
European Commission said it is "high time" AI providers ensure advanced models remain under control after new incidents involving
OpenAI agents. Meanwhile, EU Tech Commissioner Henna Virkkunen called for global rules after a report from
Anthropic unveiled the risk of AI-fueled bio weapons.
It is “high time” for artificial intelligence providers to get their house in order and ensure their advanced models are under control, a European Commission spokesperson said on Friday after news that OpenAI’s rogue AI agents affected at least 10 more websites.
The bloc has enforcement powers under the AI Act and has already sent information requests to multiple companies, it added.
“We can definitely evaluate models. We can require risk mitigation measures, and in extreme cases, and where necessary, we can also restrict, withdraw, or even recall AI models,” commission spokesperson Thomas Regnier said during a press event.
The European cybersecurity agency ENISA has access to several OpenAI models and also gained access to Anthropic’s Mythos 5 AI model, it announced on Thursday (see
here). Meanwhile, the commission has sent requests for information to more than 30 AI companies.
OpenAI and its rival Anthropic have been under regulatory scrutiny since the discovery that OpenAI’s experimental models breached Hugging Face in June. Since then, more incidents have emerged.
Earlier this week, a group of researchers discovered that OpenAI’s agents had autonomously circumvented their sandboxes and left messages on multiple websites. The incident followed last week's news that a swarm of agents hijacked the German wiki forum DseWiki sometime between May and June this year.
OpenAI’s rival Anthropic has also shared new revelations. In its latest intelligence report, published on Thursday, the company said malicious actors are using Claude to launch cyberattacks and influence campaigns (see
here). State-sponsored surveillance operations using the AI model were also discovered in China, Iran and West Africa.
One of the most serious risks, however, is the potential for biological weapons research, the company said, adding that it has blocked several such research attempts.
—Virkkunen calls for global cooperation, lawmakers look into extinction danger—
Responding to the news, the EU’s tech commissioner, Henna Virkkunen, said global AI rules are needed.
Under the AI Act, providers must assess and mitigate risks and monitor their models once they are on the market. This, however, is not the case globally, she told the press on the sidelines of a conference in Dublin.
"It's very important that we continue international co-operation,” said Virkkunen.
EU lawmakers are also reportedly looking into warnings that AI could cause human extinction by 2030.
The argument was raised by Jacob Coxon, an Anthropic researcher who previously worked at OpenAI. In his resignation letter, published on Wednesday, Coxon wrote that neither company is “acting responsibly” and is instead “racing towards super-intelligence and gambling with our lives.”
"The people building AI earnestly believe that it could kill us all by the end of the decade,” said Coxon.
Please email editors@mlex.com to contact the editorial staff regarding this story, or to submit the names of lawyers and advisers.
Tags
Sections:
Artificial Intelligence
Industries:
Computing & Information Technology
Geographies:
Europe, European Union Member States