The White House has asked OpenAI and Anthropic to withhold new AI models from the UK’s AI Security Institute (AISI). It wants US review first, according to a person familiar with the matter and a senior administration official.
The request came from the White House’s Office of the National Cyber Director, its top cybersecurity policy arm. It puts labs in the position of choosing AISI’s early access over Washington’s approval.
Anthropic Already Complies
Anthropic appears to have already fallen in line. The company withheld its newest model, Claude Mythos 5.1, from AISI’s pre-release testing. It limited initial access to a set of US organizations.
AISI has previously assessed Mythos’s advanced cyber capabilities.
Anthropic addressed the gap in a blog post announcing the model.
“We’re coordinating with the U.S. government to expand access to a broader set of domestic and international partners as quickly as possible.”
AISI Director Henry de Zoete addressed the gap in a letter to a UK parliamentary committee this month. He said the institute still has prerelease access to some of the industry’s most capable systems. De Zoete pointed to AISI’s recent review of OpenAI’s GPT-6 Astra ahead of its release.
UK Pushes Back
Since its 2023 founding, AISI has routinely tested frontier models before their public release. The institute disclosed in August that a Mythos-based agent faked identities during testing. Prime Minister Andy Burnham wants AI central to the UK’s G20 presidency next year.
UK Foreign Secretary Ed Miliband this week called for continued visibility into how AI models are tested. A UK government spokesperson said AISI continues working closely with OpenAI and Anthropic despite the access gap.
In Washington, the burden may fall on the Commerce Department’s Center for AI Standards and Innovation (CAISI). CAISI is the US body tasked with vetting frontier models before release.
CAISI currently operates without a permanent director and just a few dozen staff. Meanwhile, reports of AI systems breaching live networks during testing keep mounting. An Australian government website hack this week is the latest example.
The standoff could harden into lasting policy, or ease as CAISI scales up. Either way, the outcome may decide how much warning US allies get before the next wave of frontier models.
The post White House Wants Anthropic, OpenAI to Withhold AI Models From Britain. Why the Secrecy? appeared first on BeInCrypto.
