- The White House has reversed a September order that blocked the UK’s AI Security Institute from accessing new models from Anthropic, Google DeepMind, and OpenAI.
- The block was lifted after diplomatic pressure from UK AI Minister Kanishka Narayan.
- Anthropic had restricted UK access to its Claude Mythos 5.1 model after the September order.
- UK intelligence chiefs had separately warned ministers that AI is empowering hostile actors and terrorists.
What Happened
The White House has reversed its September decision to block the UK’s AI Security Institute (AISI) from early access to new frontier models, restoring the agency’s access to releases from Anthropic, Google DeepMind, and OpenAI, the Daily Telegraph reported, per The Decoder’s October 12, 2026 coverage. AISI tests AI systems for safety risks before they ship. Anthropic had restricted the UK’s access to its Claude Mythos 5.1 model after the original block took effect.
Why It Matters
The reversal follows direct diplomatic pressure from UK AI Minister Kanishka Narayan and caps weeks of friction between Washington and London over who gets to pre-test frontier models. It also lands just after the FTC opened a sweeping probe into OpenAI, Anthropic, and other AI labs — regulators on both sides of the Atlantic are asserting more oversight over frontier AI at the same time, even as the two governments disagree on how that oversight should work.
Technical Details
Narayan is set to announce new AI safety measures in an upcoming speech, including stricter reporting requirements for AI-related security incidents involving critical infrastructure. The UK had pushed for international coordination on AI rules at the UN General Assembly through Prime Minister Andy Burnham, while President Trump warned at the same event that “globalist” organizations should stay out of American AI development.
Who’s Affected
Anthropic, Google DeepMind, and OpenAI must now resume sharing early model access with UK regulators. The heads of Britain’s MI6, MI5, and GCHQ had separately warned ministers that AI is empowering hostile actors and terrorists, raising the stakes for AISI’s testing mandate — a concern that overlaps with Dario Amodei’s proposal to pace frontier AI releases using independent safety evaluators.
What’s Next
Watch for Narayan’s safety-measures speech and whether the restored access arrangement survives the broader US-UK disagreement over global AI governance that Trump signaled at the UN.
