Key facts
- The White House has asked OpenAI and Anthropic to delay sharing new AI models with the UK's AI Security Institute.
- The request aims to ensure US systems are secure before models are shared with international partners.
- Anthropic has not provided its Claude Mythos 5.1 model to the UK institute, stating it is available to US organizations.
- The UK's AI Security Institute has historically had privileged early access to frontier AI models.
- The US is seeking to bolster resources for the Commerce Department's Center for AI Standards and Innovation (CAISI).
The White House has requested that leading AI companies OpenAI and Anthropic refrain from sharing their newest artificial intelligence models with the UK's AI Security Institute until they have undergone testing by the US government. This directive, originating from the Office of the National Cyber Director, places these companies in a challenging position, potentially forcing them to choose between complying with the US request or risking the ire of the Trump administration.
A senior US administration official stated the White House's desire for the US to review models first to ensure American systems are secure before sharing with partners, citing company origins and existing policy. Anthropic has reportedly agreed to this request, announcing its Claude Mythos 5.1 model was exclusively available to a set of US organizations, with plans to expand access to international partners soon.
Henry de Zoete, Director of the UK's AI Security Institute (AISI), acknowledged the limited access to Anthropic's model but emphasized the institute's strong relationships with AI developers and its continued prerelease access to other capable models, including OpenAI's GPT-6 Astra. The AISI, established in 2023, has been a key part of the UK's AI safety strategy and its bid to lead global AI agreements.
Despite the potential strain on international collaboration, UK Prime Minister Andy Burnham and Foreign Secretary Ed Miliband have sought to downplay any conflict, advocating for a unified set of global principles and standards for AI safety and transparency. A UK government spokesperson reiterated the country's commitment to working closely with the US and international partners on AI testing and understanding, emphasizing that AI risks transcend national borders.
In the US, efforts are underway to increase resources for the Commerce Department's Center for AI Standards and Innovation (CAISI), considered a cornerstone of US AI governance, despite current staffing and leadership challenges.