The UK AI Security Institute (AISI) has resumed testing pre-release versions of advanced AI models from US developers. This follows a 17-day suspension of access that began on September 24, 2026, after the US Office of the National Cyber Director requested priority for domestic safety reviews.
Access was officially restored on October 11, 2026, with the institute confirming the news on October 12. Among the models currently under review is Google's Gemini 4 Argon. During the freeze, companies responded differently; for instance, Anthropic restricted access to its Claude Mythos 5.1 model for non-US entities.
UK AI Minister Kanishka Narayan played a key role in negotiating the renewed access. The AISI, established in 2023 following the AI Safety Summit at Bletchley Park, employs over 100 technical experts. The organization previously gained recognition for its evaluation of OpenAI's GPT-6 Astra.
The institute operates under voluntary agreements from 2023, which rely on developers choosing to share their systems. While the recent interruption highlighted tensions between international testing and domestic government priorities, officials stated that their relationships with AI developers remain strong.
