White House asks OpenAI and Anthropic to hold new models back from the UK AI Security Institute until the US reviews them
Politico reported on Sept 24, 2026 that the White House Office of the National Cyber Director asked OpenAI and Anthropic not to give new models to the UK's AI Security Institute for pre-release testing until US agencies had reviewed them. Anthropic appears to have complied: Claude Mythos 5.1 is available only to US organizations. The request broke with the voluntary UK pre-deployment testing that labs had used since 2023.
Key facts
- Request came from the Office of the National Cyber Director (ONCD), the White House cyber-policy office
- Anthropic: Mythos 5.1 is 'only available to a set of U.S. organizations'; 'We're coordinating with the U.S. government to expand access to a broader set of domestic and international partners as quickly as possible'
- OpenAI declined to comment; UK AISI had had pre-release access to GPT-6 Astra (The Next Web)
- Senior administration official (Politico): the US wants its systems secure before sharing with international partners, as these are American companies
- UK government: 'These risks do not stop at national borders and no country can tackle them alone'
What happened
The ONCD request, reported by Politico and followed by Bloomberg, asks the two labs to route their newest models through US government review before the UK institute sees them. Anthropic had already limited Mythos 5.1 to US organizations. The UK government answered with a general call for international cooperation. The same week, reports said UK AISI staff were under strain from tight release schedules (FT, Sept 22).
Why it matters
Independent pre-deployment testing by the UK institute was one of the few working international safety mechanisms. Putting a US gate in front of it treats frontier models as national-security assets and weakens the "international testing" pillar that the labs themselves proposed at the UN the same week.
Caveat: based on anonymous sources in Politico; the exact terms of the request and OpenAI's decision were not public.
Changelog
- 2026-09-29: created (sweep 2026-09-29)
Related events
- Anthropic releases Claude Fable 5.1 and Claude Mythos 5.1 ★★★★★
- US government asks OpenAI to limit GPT-5.6 release to approved partners ★★★★
- Trump at the UN General Assembly 'totally rejects' any global scheme to control AI and renames it 'super intelligence' ★★★★
- UK PM Andy Burnham says the UK will use its 2027 G20 presidency to broker a global AI agreement ★★
Sources (3)
- pressPolitico: White House asks OpenAI and Anthropic to hold new models from UK testers until U.S. review
- pressThe Next Web: White House wants US review before UK AI Security Institute tests
- pressDigital Watch Observatory: US asked OpenAI, Anthropic to hold AI models from UK
id: 2026-09-24-white-house-asks-labs-withhold-models-uk-aisi · updated 2026-09-29 · open in the interactive timeline