White House reportedly reverses request to withhold new AI models from UK testers

The Telegraph reported on October 11, 2026, that the White House reversed its earlier request to withhold new AI models from the UK's AI Security Institute until US testing. It says lobbying by UK AI minister Kanishka Narayan preceded the change, and the institute will receive prerelease access to models from Anthropic, OpenAI and Google DeepMind. POLITICO reported in September that Anthropic had not provided Claude Mythos 5.1 to the institute.