The White House has asked OpenAI and Anthropic to subject their newest AI models
to a U.S. government security review before granting testing access to the UK AI
Safety Institute (AISI), Politico reports. The request was made by the Office of
the National Cyber Director; the administration wants U.S. authorities to assess
frontier models first and clear them for allied access only after U.S. systems
are deemed secure. Anthropic has not provided Claude Mythos 5.1 to the UK AISI
and says it is coordinating with the U.S. government to phase access for
domestic and foreign partners. The development comes as Washington debates
regulation of advanced AI and moves to strengthen the Commerce Department’s
Center for AI Standards and Innovation (CAISI); OpenAI has urged CAISI to work
with international bodies including the UK AISI on global AI safety standards.
