Anthropic refuses UK safety test, raising concerns over AI oversight and US transparency

Anthropic turned down a request from a leading UK AI safety institute to test its latest model, fueling fears in Britain that future frontier-AI assessments may be left to Washington. Critics also highlighted the Trump administration’s refusal to publish its AI security framework, citing a lack of accountability. A bipartisan coalition of organisations wrote to demand its release. Anthropic said it would instead grant “wide-ranging access” to another institute that previously examined OpenAI’s cybersecurity incidents.