White House asks OpenAI and Anthropic to hold new models from UK testers until US review

The White House has asked OpenAI and Anthropic to hold back new AI models from the United Kingdom's AI Safety Institute (AISI) until U.S. agencies review them first. Politico reported the request on September 24; it came from the Office of the National Cyber Director.

AISI is one of the best-equipped testing bodies in the world

The British institute is considered one of the world's best-equipped government AI testing bodies and has early access to frontier models from leading labs. It recently became the first agency to report autonomous deception by AI agents in real-world settings.

In a letter to Parliament, AISI Director Henry de Zoete said the institute still has access to frontier models and tested OpenAI's GPT-6 Astra before its release.

Anthropic falls in line

Anthropic has already complied: its latest model, Claude Mythos 5.1, is available only to U.S. organizations. According to Politico, both companies had to choose between withholding models from AISI and angering the White House.

Washington's own capacity problem

The U.S. body responsible for AI standards, the Center for AI Standards and Innovation (CAISI), has problems of its own: it has no permanent director and only a few dozen staffers. That leaves open the question of how quickly Washington can take over the testing AISI has been doing. British Prime Minister Andy Burnham played down the conflict at the UN General Assembly and called for "shared global principles and standards."

First access to frontier models, once a routine part of allied safety cooperation, has become a matter of national priority. For the labs, decisions that looked technical now carry a political cost.