The White House has requested that OpenAI and Anthropic restrict the UK’s AI Security Institute (AISI) from accessing their newest AI models. The directive, issued by the Office of the National Cyber Director, aims to prioritize US government review before any international release.
Anthropic has already complied with this request by withholding its latest model, Claude Mythos 5.1, from the institute’s pre-release testing protocols. The company currently limits initial access to a select group of US-based organizations while coordinating with the US government to expand access to international partners.
Despite these restrictions, AISI Director Henry de Zoete confirmed that the institute maintains pre-release access to other advanced systems, including a recent review of OpenAI’s GPT-6 Astra. The UK government continues to advocate for transparency in AI testing, with officials emphasizing ongoing collaboration with both companies.
In the United States, the responsibility for vetting frontier models rests with the Commerce Department’s Center for AI Standards and Innovation (CAISI). The agency is currently operating with a limited staff of several dozen employees and lacks a permanent director, which may influence future policy decisions regarding international AI model access.
