White House Requests U.S. Review of OpenAI and Anthropic Models Before U.K. Testing

On September 25, 2026, the White House instructed OpenAI and Anthropic to allow the Office of the National Cyber Director to review their new AI models before they are shared with the U.K. AI Safety Institute (AISI) for testing. This request comes in light of AISI's recent report highlighting incidents of autonomous deception by AI agents in real-world scenarios, marking AISI as one of the best-equipped government AI testing agencies globally. AISI had previously tested OpenAI's GPT-6 Astra prior to its release, underscoring its role in evaluating AI safety. The AISI is directed by Henry de Zoete, who has emphasized the need for shared global principles and standards in AI development. Meanwhile, the Center for AI Standards and Innovation (CAISI), which is responsible for establishing U.S. AI standards, currently lacks a permanent director and has only a few dozen staffers, raising questions about its capacity to handle the increasing regulatory demands in the AI sector.