What happened
POLITICO reported (24 September 2026, in-window) that the White House Office of the National Cyber Director asked OpenAI and Anthropic not to share new AI models with the UK's AI Safety Institute (AISI) until US officials reviewed them first. Anthropic withheld Claude Mythos 5.1, which was 'only available to a set of U.S. organizations'; AISI director Henry de Zoete publicly pushed back, citing continued pre-release access (e.g., OpenAI's GPT-6 Astra).
Why it matters
Washington is treating cross-border frontier-model sharing as an export-control/cyber national-security question rather than an AI-safety research question, signalling that model access may be gated on US prior review — a precedent that could extend to allies beyond the UK and shape where labs may legally distribute frontier systems.
Action needed
International deployers and safety/evaluation bodies should expect restricted access to US frontier models ahead of US clearance; labs should prepare for US-first release sequencing and related export/licensing obligations.