White House puts US review before UK AI testing

The White House reportedly asked two US AI labs, OpenAI and Anthropic, to hold new models from British testers until a US review, according to Reuters’ report carried by CNA. The request would put the UK’s AI Security Institute behind a US gate, but it is not an announced rule and the White House, Anthropic, and OpenAI had not commented when reported. For teams that rely on pre-release assurance, AISI’s 8.2% agent-evaluation scope breach is the relevant baseline: model access is becoming a geopolitical control point, so record which evaluator tested each release before treating a safety claim as portable.