White House Restores UK AI Safety Institute Access to Frontier Models After 17-Day Freeze, Following Diplomatic Pressure

The White House lifted a 17-day freeze on October 11, restoring the UK AI Safety Institute's access to pre-release models from OpenAI, Anthropic, and Google DeepMind, according to The Daily Telegraph and Reuters. The Office of the National Cyber Director had imposed the pause on September 24, telling the companies to hold new models back from British testers until Washington finished its own national security review, with a senior U.S. administration official telling City A.M. it reflected policy that American companies provide access to U.S. reviewers first. Anthropic complied by withholding Claude Mythos 5.1 from the UK institute during the freeze, keeping the model with a set of U.S. organizations.

The reversal followed lobbying by UK AI Minister Kanishka Narayan, according to multiple reports. The UK AI Safety Institute, established after the 2023 AI safety summit, evaluates frontier AI systems for safety risks before public release under voluntary agreements struck with developers, meaning its access depends on companies choosing to share their models and is now explicitly subject to U.S. government direction. The 17-day episode demonstrates that pre-release testing access, long treated as a background courtesy between U.S. labs and allied regulators, has become a lever of national AI policy that Washington can withdraw and restore based on its own security review timeline, creating uncertainty for non-U.S. safety institutes that rely on voluntary sharing arrangements with American frontier labs.

Sources

The Daily Telegraph, UK AI Safety Institute access restored, October 11-12, 2026, cited by The Decoder at the-decoder.com and AIWeekly at aiweekly.co and Startup Fortune at startupfortune.com Reuters, UK given access to safety-test latest AI models, minister says, October 12, 2026, whtc.com and cited at startupfortune.com City A.M., US administration official statement on policy, September-October 2026, cited by AIWeekly at aiweekly.co AIWeekly, White House restores UK AISI access after 17-day model freeze, October 12, 2026, aiweekly.co The Decoder, White House reverses course and restores UK AI Safety Institute's access to frontier models, October 12, 2026, the-decoder.com CryptoBriefing, US reverses course, restores UK access to frontier AI models for testing, October 11, 2026, cryptobriefing.com Startup Fortune, UK regains access to AI models after White House lifts 17-day freeze, October 12, 2026, startupfortune.com