Stories
White House Restores UK Access to Advanced AI Models for Safety Checks
- •The White House has reversed its previous restriction, allowing the UK to regain access to advanced American artificial intelligence models for pre-release safety testing.
- •British officials will now be able to examine unreleased systems from developers such as OpenAI, Anthropic and Google DeepMind via the UK's AI Security Institute.
- •The policy change follows intense lobbying by British ministers, including AI minister Kanishka Narayan, after earlier interventions disrupted transatlantic testing arrangements.
- •The UK is also preparing new cyber security measures, including updates to the Cyber Security and Resilience Bill, to strengthen defences against AI-related threats.
Key Questions Answered
Following lobbying by British ministers like AI minister Kanishka Narayan, the White House withdrew its request for leading developers to withhold unreleased systems from the UK's AI Security Institute. This reversal removes obstacles to transatlantic cooperation on AI safety and allows British officials to examine models from OpenAI, Anthropic, and Google DeepMind before public release.
The dispute emerged when the White House instructed major American AI companies to stop providing the UK's AI Security Institute with access to new models ahead of their release. Consequently, Anthropic restricted British officials from examining its Claude Mythos 5.1 model.
Ministers are preparing changes to Labour’s Cyber Security and Resilience Bill to strengthen reporting obligations for critical national infrastructure organizations during AI-related security incidents. The government is also considering wider use of advanced AI tools to protect Whitehall departments and providing additional resources to the National Cyber Security Centre.
