White House asks OpenAI and Anthropic to delay UK model access

By Splitfeed EditorialLast updated Sep 26, 2026, 19:17 UTC

On September 24, 2026, Politico reported that the White House asked both OpenAI and Anthropic to withhold their upcoming AI models from the United Kingdom's AI Safety Institute until US government agencies have completed their own review first. The request names both companies explicitly and applies to new models, meaning British safety testers would be placed behind US reviewers in the access queue. The UK AI Safety Institute is the British government body established to evaluate frontier AI models for potential safety risks, and it had previously been granted access to models from US developers as part of a bilateral arrangement.

The move matters because it signals a meaningful shift in how the US government is treating AI safety cooperation with close allies. The prior arrangement with the UK was seen as a model for international coordination on frontier AI oversight, and placing US review ahead of British access effectively subordinates that partnership to domestic gatekeeping. What remains in dispute is the legal or policy basis on which the White House made the request, whether OpenAI and Anthropic agreed to comply, and whether this reflects a broader recalibration of US AI diplomacy or a narrower procedural adjustment.

The anti-AI camp has reacted with a mix of frustration and skepticism, though not in the direction one might expect. Accounts like tweety fish argued that alignment-style safety risks do not actually exist in large language models and that companies like OpenAI and Anthropic have a commercial interest in promoting them, suggesting the entire safety-testing framework is built on a shaky premise. Scott, posting from gatorgrad.bsky.social, directed pointed criticism at the White House decision-making itself. Mrs. Betty Bowers offered a broader swipe at AI industry terminology. The pro-AI camp and the middle ground camp have not yet posted reactions to this specific story. The pro-AI camp would typically argue that US review should come first to protect competitive and national security interests, while the middle ground camp would likely ask whether delaying allied access undermines the cooperative safety infrastructure both governments have invested in building.

The clearest signal to watch for is whether OpenAI or Anthropic publicly confirm compliance with the request, and whether the UK AI Safety Institute or the British government responds formally, which could reveal how much strain this places on the bilateral AI safety relationship established in recent years.