White House restores UK AISI access after 17-day model freeze

Britain's AI Security Institute will again get pre-release access to frontier models from OpenAI, Anthropic and Google DeepMind, the Telegraph reports, after the White House lifted a 17-day freeze on October 11.

The pause began on September 24, when the Office of the National Cyber Director told the labs to hold new models back from British testers until Washington finished its own national security review. 'Because they're American companies and this has been our policy with every new frontier model that comes out,' a senior US administration official told City A.M.

Anthropic complied first and most visibly. The company did not share Claude Mythos 5.1 with AISI during the freeze, keeping the model with a set of US organizations. 'We're coordinating with the US government to expand access to a broader set of domestic and international partners as quickly as possible,' Anthropic said. The reversal followed lobbying by AI minister Kanishka Narayan.

AISI director Henry de Zoete acknowledged his institute had not received Anthropic's latest model while the freeze was in force, but said AISI maintains 'strong relationships with all frontier AI developers,' noting it had tested OpenAI's GPT-6 Astra before its release. The access squeeze arrived at a bruising internal moment: AISI chief technology officer Jade Leung is stepping back from her full-time role at month's end, remaining as vice-chair.

It is one of a run of Anthropic items on our tracker this week where the story is about Washington's hand on the valve rather than about the model itself.