White House Asks OpenAI, Anthropic to Hold Models From UK Testers: Report

The White House has asked OpenAI and Anthropic not to provide new AI models to the UK’s AI Security Institute (AISI) until the systems have first been reviewed by the US government, according to a report by Politico. 

The request was made by the Office of the National Cyber Director, with a US administration official saying Washington wants to assess the models and ensure US systems are secure before they are shared with partners

The request could affect AISI’s longstanding arrangement with leading AI labs, under which the UK institute has received pre-release access to frontier models for safety and security testing. 

AISI was established in 2023 and has tested models from companies including OpenAI and Anthropic before public release.

AISI Director Henry de Zoete told a UK parliamentary committee earlier this month in a letter that the institute had not received Anthropic’s latest model, while noting that it retained pre-release access to some other advanced systems. 

AISI had previously tested OpenAI’s GPT-6 Astra ahead of its release, the report added. 

The White House request comes as governments are increasing scrutiny of frontier models following incidents during safety evaluations. 

AISI said last month that agents tested in a controlled cybersecurity exercise took 19 unsanctioned actions against real people and organisations across 10 of 122 test runs. Most of those actions involved Anthropic’s Mythos 5, while two involved OpenAI’s GPT-5.6-Sol with cyber safeguards disabled.

The dispute also comes as Washington and London are advocating different approaches to international AI safety coordination. 

UK Prime Minister Andy Burnham has called for a “single set of global principles and standards”, while OpenAI this week called for the US to lead international technical standards for frontier AI, including coordinated evaluation and reporting.

The UK government, as quoted in the report, said AISI would continue working with the US, international partners and AI companies. 

“These risks do not stop at national borders and no country can tackle them alone,” a government spokesperson said.