The U.S. government is asking major artificial intelligence developers to give American regulators first access to new models. According to reporting from The Decoder, the White House has instructed OpenAI and Anthropic to prioritize U.S. government evaluation before sharing upcoming models with the United Kingdom’s AI Safety Institute. The request marks a notable shift in how the U.S. coordinates AI safety evaluations with international allies.
Geopolitical and Regulatory Context
Both OpenAI and Anthropic have previously committed to voluntary safety frameworks, including sharing pre-release models with safety institutes to evaluate potential risks such as cybersecurity vulnerabilities or dangerous capabilities. While the U.S. and the U.K. have established formal partnerships to share information regarding artificial intelligence oversight, this request suggests Washington wants domestic federal agencies to conduct early inspections prior to foreign government review. Placing American regulatory priority over international partners highlights growing national security considerations surrounding advanced frontier models developed by U.S.-based companies.
Key Details Unclear
The reported instructions leave several key details open. The initial report did not specify which specific U.S. agencies will handle the evaluations, nor did it clarify whether the policy applies strictly to OpenAI and Anthropic or if it will extend to other American AI firms. Furthermore, details regarding how long U.S. officials will retain exclusive access before British testers receive the models remain undisclosed. It is also unclear how this directive impacts existing bilateral AI safety agreements between the U.S. and the U.K., or whether British officials were consulted before the direction was issued to the companies.
As AI development accelerates, national governments are increasingly treating model safety testing as a priority linked to national security. How OpenAI, Anthropic, and international safety institutes adapt to this national review sequence will shape future global governance efforts.
Lead image: RDNE Stock project / Pexels
