Why OpenAI Is Letting the U.S. Government Examine New ChatGPT Models

OpenAI has confirmed that it will voluntarily align with a new artificial intelligence executive order introduced by U.S. President Donald Trump, as stated by the company’s Head of Countries, George Osborne. The directive aims to introduce a structured review system for advanced AI technologies, requiring developers to provide access to their newest models 30 days prior to public release. While the order does not legally compel companies to submit their systems, it encourages transparency through early evaluation. Earlier drafts of the policy had proposed a stricter framework, including mandatory submission of AI models up to 90 days before launch, but those provisions were later revised following industry pushback and political concerns. According to Osborne, OpenAI will participate in the proposed benchmarking process, which is designed to assess high-level capabilities such as cybersecurity risks and advanced system behavior. The framework will also determine whether an AI system qualifies as a “frontier model,” a classification reserved for highly sophisticated technologies requiring additional oversight. Speaking at a tech event in London, Osborne noted that OpenAI has already engaged with governments worldwide to discuss AI safety, security, and responsible deployment. He emphasized that these discussions extend beyond the United States and are part of a broader global effort to manage emerging AI risks.