The Trump administration has formally requested that OpenAI limit the initial release of its upcoming GPT-5.6 model to a select group of government-vetted partners. This marks an unprecedented intervention by the U.S. government into an American AI company's product launch, driven by security concerns and the advanced capabilities of the new model. The White House's Office of the National Cyber Director and Office of Science and Technology Policy initiated this request as the administration works to establish a framework for testing and evaluating the security of advanced AI models.
OpenAI CEO Sam Altman confirmed this limited rollout in an internal memo, stating that while this isn't their preferred long-term approach, they are working with the government to achieve a more sustainable model for future releases. The request comes after the Commerce Department previously restricted access to Anthropic's advanced models, Fable 5 and Mythos 5, over similar security fears. Sources indicate that GPT-5.6 possesses "Mythos-like" capabilities, prompting the government's caution.
This move by the government is a direct response to a recent executive order signed by President Trump, which mandates voluntary testing protocols for AI companies before releasing new models. Although the framework for this order is still under development, the administration's actions signal a growing intent to exercise oversight over frontier AI development. The situation highlights the delicate balance AI labs face in competing to release new models while navigating emerging regulatory pressures and concerns about national security, especially when compared to the rapid advancement of Chinese open-source AI models.
The initial plan for OpenAI was to provide early access to approximately 24 partners. However, the White House, in coordination with the Departments of Commerce and Treasury, pushed to narrow this list and subject it to government approval. This strategy transforms "early access" into a permissioned tier, where early adopters are screened and a staged release becomes the norm, emphasizing that Washington views frontier AI as a strategically sensitive capability rather than a typical software launch.
