White House is asking OpenAI and Anthropic to delay sharing new AI models with U.K. testers

The White House has asked OpenAI and Anthropic to hold new AI models from British safety testers pending a U.S. review, according to a Politico report published Thursday.

The administration wants to ensure U.S. AI systems are secure before models are shared with partners, according to Reuters. The White House, Anthropic, and OpenAI did not respond to Reuters' requests for comment.

The decision reflects growing White House concern over cybersecurity vulnerabilities in increasingly capable AI models, amid a string of incidents in which AI systems have broken into real-world computer systems without authorization, according to Reuters. On Wednesday, Australian officials revealed that in June an OpenAI agent broke into a government health data portal and obtained unauthorized access to files.

Both companies have separately raised alarms at the United Nations Security Council, calling on governments to coordinate their approach to managing the dangers that advanced AI systems present, according to Reuters.

The request adds a new dimension to broader conversations within the AI industry about safety and international coordination. OpenAI, Anthropic, and Google $GOOGL DeepMind — a unit of Alphabet — have been engaging on AI safety for several weeks, with OpenAI's global policy chief Chris Lehane arguing that no antitrust waiver is needed for the companies to collaborate on safety. Anthropic CEO Dario Amodei had published an open letter calling on AI companies and governments to slow the pace of frontier model development.

Senior representatives from Nvidia $NVDA, OpenAI, Anthropic, and Google DeepMind gathered at Dumfries House in Scotland on Thursday at a summit hosted by King Charles III focused on AI safety and the technology's societal role. U.K. AI Minister Kanishka Narayan also attended the event, which sought to build agreement on common guiding principles for AI development.