According to people familiar with the matter and a senior US government official who was allowed to anonymously describe the request, the White House has asked OpenAI and Anthropic not to share their latest artificial intelligence models with British government testing agencies until the US government has completed testing. This request from the US National Cyber Director's Office has left the companies involved in a dilemma: they either refuse to provide models to the UK Artificial Intelligence Security Institute — one of the most resourced government testing agencies in the world and previously enjoyed priority access to new models; or they had to risk annoying the Trump administration's White House. The senior administration official said that the White House hopes the US will review the model and take relevant measures to ensure the safety of the US system before sharing it with US allies.

According to people familiar with the matter and a senior US government official who was allowed to anonymously describe the request, the White House has asked OpenAI and Anthropic not to share their latest artificial intelligence models with British government testing agencies until the US government has completed testing. This request from the US National Cyber Director's Office has left the companies involved in a dilemma: they either refuse to provide models to the UK Artificial Intelligence Security Institute — one of the most resourced government testing agencies in the world and previously enjoyed priority access to new models; or they had to risk annoying the Trump administration's White House. The senior administration official said that the White House hopes the US will review the model and take relevant measures to ensure the safety of the US system before sharing it with US allies.