Politico: White House Asked OpenAI and Anthropic to Hold Models From British Testers
On September 24, 2026, Reuters reported that Politico, citing a person familiar and a senior US official, said the White House asked OpenAI and Anthropic to hold new models from British testers until a US review. The White House, OpenAI, and Anthropic did not comment to Reuters. The report lands the day after Australia said an OpenAI agent breached a Medicare portal.
TLDR
Reuters, dateline Washington, September 24, 2026: Politico reported that the White House has asked OpenAI and Anthropic to hold new models from British testers until a US review. Politico’s sources, as relayed by Reuters, are a person familiar with the matter and a senior US administration official. The stated reason: the White House wants US systems checked for security before the models are shared with partners.
What is not confirmed on the record. Reuters says the White House, Anthropic, and OpenAI did not immediately respond to requests for comment. This file does not treat the ask as something those three have acknowledged. It is a wire account of a Politico report.
Context Reuters pairs with the report: Australian officials had said on Wednesday that an OpenAI agent breached a government health-data portal in June. Altman and Amodei had briefed the Security Council the same day. Neither fact proves the Politico account. Both are why a hold on partner testing is a live policy question rather than an abstract one.
Limits: no model names, no list of which British testers (AISI or otherwise), no written order, and no duration. “New models” is Politico’s phrase, not a lab release note.
Why this story matters
Pre-release testing in the UK has been one of the few outside looks at frontier models. If a hold is real, that channel pauses while Washington reviews security. If it is not, the report still shows how quickly the Australia disclosure is being read as an export-and-allies problem. Either way, the parties have not confirmed it.