Anthropic declined to submit its latest AI model to a leading UK safety institute, a move that has prompted concern in the United Kingdom that oversight of frontier AI systems could become concentrated in the United States. British observers worry that if companies do not engage with UK reviewers, reviews of state-of-the-art systems may be dominated by Washington-based processes.
Growing transparency concerns over US AI security framework
At the same time, questions are mounting about the lack of transparency surrounding the Trump administration’s AI security framework, which has not been made public. Prominent organisations from across the political spectrum have requested its release in a letter first shared with Semafor. The letter warned: “A lack of accountability over the governance of a [critical] technology would… likely lead to a loss of public trust.”
Anthropic says it will give alternative access
Anthropic said on Wednesday that it would provide “wide-ranging access” to an institute that previously conducted a review into cybersecurity incidents at OpenAI. The company’s statement did not detail the exact terms or scope of that access in public disclosures.
Why this matters
The episode matters because independent review and transparency around high-capability AI models are central to public trust and international cooperation. If evaluation of advanced systems is handled primarily by institutions in one country or region, it could complicate global regulatory efforts and undermine confidence among other governments and stakeholders.
Further developments — including any response from the UK institute or clarification from the US government about the unpublished framework — will be important in shaping international AI safety collaboration.
