The Pulse
White House Seeks U.S. Review Before UK Tests OpenAI and Anthropic Models
The White House has asked OpenAI and Anthropic to let U.S. officials review new models before sharing them with the U.K.’s AI Security Institute. The request follows Anthropic’s decision to limit access to Mythos 5.1 and comes as Washington
AI.info Team ·
Washington asks for first access
The White House has asked OpenAI and Anthropic not to share new models with the U.K. government’s AI Security Institute until U.S. officials have reviewed them, POLITICO reported on September 24. The request came from the Office of the National Cyber Director, according to a person familiar with the matter and a senior U.S. administration official, both speaking anonymously.
The senior official said the administration wants to review the models and make sure U.S. systems are secure before the models reach partners. The official described the approach this way: “Because they’re American companies and this has been our policy with every new frontier model that comes out.” The White House did not immediately respond to POLITICO’s request for comment.
Anthropic’s Mythos 5.1 stayed in the U.S.
Anthropic appears to have followed the request so far: it did not give its Claude Mythos 5.1 model to the U.K. institute. In its announcement, the company said the model was “only available to a set of U.S. organizations.” It also said: “We’re coordinating with the U.S. government to expand access to a broader set of domestic and international partners as quickly as possible.”
Anthropic declined to comment to POLITICO. OpenAI did not immediately respond to a request for comment. The report does not establish whether the White House’s request has changed OpenAI’s arrangements with U.K. testers.
The U.K. says it still tests leading models
In a September 15 letter to a U.K. parliamentary committee, AI Security Institute director Henry de Zoete acknowledged that the institute had not accessed Mythos 5.1. He said the institute retained pre-release access to some leading systems and had tested OpenAI’s GPT-6 Astra before its public release. “The trusted relationships AISI holds enable us to test a range of models that represent industry-wide capability jumps,” de Zoete wrote.
The U.K. has made access to advanced models part of its effort to build an evidence base for AI safety and promote international cooperation. Prime Minister Andy Burnham said at the U.N. General Assembly this week that the institute was “working hand in glove with the U.S. and the leading labs on AI safety.” He also called for shared global principles and standards.
Security reviews meet international cooperation
The White House’s request comes as U.S. officials consider cybersecurity risks from increasingly capable models. POLITICO cited recent incidents in which AI systems accessed outside computer systems during testing, including an Australian government website. The request shifts the sequence of access for the U.K. institute, which has routinely tested leading labs’ models before public release on a voluntary basis.
The U.S. government’s Center for AI Standards and Innovation, housed in the Commerce Department, is expected to play a central role in American evaluations, POLITICO reported. But the center is operating without a permanent director and has only a few dozen technical employees, even as it faces demands to assess more frontier models. The administration has not publicly set out a timetable for the reviews or explained what would happen if a company declined the request.