Glonce

White House reportedly restores UK AI Security Institute's early access to frontier models

policyThe Glonce editorsThe Decoder🔥55Glonce rating55Impact55Novelty40Relevance75Surprise40Players55Credibility65
White House reportedly restores UK AI Security Institute's early access to frontier models

The White House has reversed its decision to block the UK's AI Security Institute (AISI) from accessing new AI models, according to The Daily Telegraph as relayed by The Decoder. AISI tests AI systems for safety risks before they ship, and it will now once again get early access to models from Anthropic, Google DeepMind and OpenAI.

The block dates from September, when the Trump administration told the companies to stop sharing their models with AISI. Anthropic then restricted access to its new Claude Mythos 5.1 model. The block has now been lifted after diplomatic pressure from the UK's AI Minister, Kanishka Narayan.

Narayan plans to announce new AI safety measures in an upcoming speech. They include stricter reporting requirements for AI-related security incidents involving critical infrastructure.

The article also sets out the political backdrop. Prime Minister Andy Burnham had previously pushed for international AI regulation at the UN General Assembly, and Trump warned at the same event that "globalist" organizations should stay out of American AI development. The heads of MI6, MI5 and GCHQ had also warned ministers that AI was empowering hostile actors and terrorists.

Key facts

  • The White House has reversed its decision to block the UK's AI Security Institute (AISI) from accessing new AI models, per The Daily Telegraph.
  • AISI will once again get early access to models from Anthropic, Google DeepMind and OpenAI.
  • The block began in September, when the Trump administration told the companies to stop sharing models with AISI; Anthropic then restricted access to its new Claude Mythos 5.1 model.
  • The block was lifted after diplomatic pressure from AI Minister Kanishka Narayan, who plans to announce new AI safety measures in an upcoming speech.
  • The planned measures include stricter reporting requirements for AI-related security incidents involving critical infrastructure.

Why it matters

AISI's job is to test AI systems for safety risks before they ship, and that depends on getting the models early. The September instruction to the three companies to stop sharing models cut off that access, and Anthropic then restricted access to its new Claude Mythos 5.1 model. Restoring early access to models from Anthropic, Google DeepMind and OpenAI puts pre-release testing by the UK institute back on the table. The story also sits against a tense political backdrop: Prime Minister Andy Burnham had pushed for international AI regulation at the UN General Assembly, where Trump warned that "globalist" organizations should stay out of American AI development.

Who it affects

Most directly the UK AI Security Institute, which tests frontier systems, and the three companies named: Anthropic, Google DeepMind and OpenAI, whose models it is due to receive early. It also affects the UK government, with AI Minister Kanishka Narayan credited with the diplomatic pressure behind the reversal. The safety measures Narayan plans to announce include stricter reporting requirements for AI-related security incidents involving critical infrastructure.

How to use it

There is nothing to use directly; this is a government and access story, not a product release. The thing to watch is Narayan's upcoming speech, where she is reported to be set to announce new AI safety measures, including stricter reporting requirements for AI-related security incidents involving critical infrastructure.

How solid is it

Moderately. The reversal is reported by The Daily Telegraph and relayed by The Decoder, and the source does not say The Decoder confirmed it independently. No statement from the White House, AISI, Anthropic, Google DeepMind or OpenAI is quoted. The September block, Anthropic's restriction on Claude Mythos 5.1 and Narayan's role are stated by The Decoder's article without a quoted primary source.

Risks and caveats

Key details are missing. The source gives no year for the September instruction and no date for the reversal, and it offers no reason for why the White House first blocked AISI's access or why it reversed. It is not said whether Anthropic has lifted its restriction on Claude Mythos 5.1 for AISI. No date or venue is given for Narayan's speech. The source also says only that the political events (the UN General Assembly push and Trump's remark) came earlier; it does not say they caused the block or the reversal. Its headline calls the body the AI Safety Institute while the text says AI Security Institute, and it does not explain the difference.