White House asks OpenAI, Anthropic to hold models from British testers, Politico reports - Reuters

The White House has requested that major artificial intelligence developers, specifically OpenAI and Anthropic, refrain from sharing their newest models with the United Kingdom’s testing agency until a formal review by the United States government is completed. This directive aims to ensure that American authorities have the opportunity to evaluate the security and safety implications of advanced AI systems before they are subjected to international scrutiny.

The request reflects an increasing focus by the U.S. government on the oversight of frontier AI development. By prioritizing domestic review, officials seek to maintain a level of control over the deployment and testing of technologies that could have significant national security or safety consequences. This approach suggests a desire to align international testing protocols with U.S. regulatory standards.

The move highlights the growing complexity of international cooperation regarding AI governance. While the United Kingdom has established its own dedicated agency to evaluate AI risks, the U.S. government is asserting its interest in reviewing these models first. This creates a potential bottleneck for developers who are currently working to satisfy both American and British regulatory requirements.

Industry observers note that this request places OpenAI and Anthropic in a delicate position as they navigate varying international expectations. The companies are currently at the forefront of AI innovation and are subject to intense pressure to demonstrate the safety of their products. How these developers balance the demands of different national governments remains a key point of interest for the technology sector.

The situation underscores the broader challenge of creating a unified global framework for AI safety. As nations move to establish their own testing and evaluation bodies, the potential for conflicting regulatory requirements increases. The U.S. government's intervention indicates that it intends to remain a central authority in the evaluation process for the most powerful AI models currently in development.