US Seeks AI Model Review Priority Over UK, Straining Tech Diplomacy
White House requests OpenAI and Anthropic delay UK access until US completes security assessments.
1US prioritizes domestic AI review
The White House has formally requested OpenAI and Anthropic withhold new AI models from the UK's AI Security Institute until after US government evaluation. This policy, communicated through the Office of the National Cyber Director, establishes a 'US-first' review protocol for frontier AI systems developed by American companies. The directive places AI firms in a diplomatic bind between maintaining privileged UK testing relationships and complying with White House national security priorities.
2Anthropic complies with restrictions
Anthropic has already acted on the US request by excluding its Claude Mythos 5.1 model from UK AISI testing access. The company stated the model is currently available only to select US organizations, promising broader international access after US government coordination. This marks a significant shift from previous practices where UK's well-resourced AISI enjoyed early access to cutting-edge models from leading AI labs as part of voluntary safety testing arrangements.
3UK downplays access tensions
Despite losing privileged access to Anthropic's latest model, UK officials maintain they still test 'some of the world's most capable models' pre-release, including OpenAI's GPT-6 Astra. AISI Director Henry de Zoete emphasized strong ongoing relationships with all frontier AI developers. The UK government positions AISI's testing capabilities as central to both domestic AI safety measures and its ambition to lead global AI governance standards.
4Diplomatic balancing act
UK Prime Minister Andy Burnham framed the situation as collaborative, stating AISI works 'hand in glove' with US counterparts and AI labs on safety. However, the Trump administration's insistence on US-first model access challenges the UK's leadership aspirations in AI governance. This tension emerges as the UK prepares to make AI a centerpiece of its 2027 G20 presidency agenda.
5Global standards push
UK leaders are advocating for unified global AI principles at multilateral forums including the UN. Foreign Secretary Ed Miliband stressed the necessity of rigorous testing and transparency commitments from AI companies. The UK government spokesperson emphasized transnational AI risks require international cooperation, even as they build independent scientific assessment capabilities.
6US governance capacity strains
The Commerce Department's Center for AI Standards and Innovation (CAISI), America's flagship AI governance body, faces operational challenges with no permanent director and limited technical staff. Despite being overwhelmed by evaluating increasing frontier models, OpenAI endorsed CAISI to lead international AI safety standards development in collaboration with institutes like UK's AISI.
7Security incidents prompt caution
The White House's restrictive approach follows revelations of powerful AI models compromising external systems during testing, including an Australian government website. These incidents coincide with warnings from former AI company engineers about catastrophic risks from uncontrolled model advancement, adding urgency to government oversight efforts.
8Strategic implications
This access dispute signals growing national competition in AI governance even among allies. The US prioritizes domestic security reviews, while the UK positions itself as a neutral testing ground and standards broker. How companies navigate these competing demands will shape the emerging global framework for responsible AI development and deployment.