White House wants US government to review OpenAI, Anthropic’s new AI models before UK entry — Report | Today’s news
The White House has asked OpenAI and Anthropic not to share their latest AI models with the U.K.’s artificial intelligence testing agency until the models first pass testing by the U.S. government, Politico reported.
The request was reportedly made by the Office of the National Cyber Director. A senior US administration official said the White House wants to review the models and take steps to ensure the security of US systems before they are shared with US partners, according to the newspaper.
“Because they are American companies and this has been our policy with every new frontier model that comes out,” the official said.
Antropic holds the latest model from the UK
So far, Anthropic appears to have complied with the request. The company did not provide its Claude Mythos 5.1 AISI model, saying that the model is “only available to a group of US organizations”.
The development marks a change for the British institute, which has built relationships with major AI developers and routinely tests advanced models before their public release on a voluntary basis.
AISI director Henry de Zoete admitted the lack of access to the latest Anthropic model in a letter to a UK parliamentary committee earlier this month.
However, he said the institute continues to have preliminary access to some of the world’s most capable AI models, noting that it tested the OpenAI GPT-6 Astra before its release.
“The trusted relationships that AISI maintains allow us to test a number of models that represent leaps in capability across the industry,” de Zoete wrote.
AI cybersecurity concerns behind US review
The White House’s request comes as the Trump administration considers how to handle increasingly powerful artificial intelligence border systems.
Recent testing has shown that some advanced models can perform sophisticated cyber activities, including hacking into external systems. An Australian government website was among the systems reportedly targeted during the testing, adding to concerns about the potential security risks posed by increasingly capable AI.
The US administration’s push for an initial domestic review could therefore reflect concerns about the security implications of giving foreign governments access to powerful models before US authorities review them.
The UK is trying to ease tensions
The development comes as the UK seeks to position itself as a leader in international security and AI testing.
Created in 2023, AISI has made access to frontier models and its testing expertise an important part of the UK’s approach to AI security. The UK has also encouraged international cooperation on common AI security principles and standards.
“These risks do not stop at national borders and no country can tackle them alone,” a UK government spokesman said, adding that Britain would continue to work with the US and other partners to test advanced artificial intelligence systems.
