Trump Administration Seeks OpenAI Limits on GPT-5.6

White House Seeks to Limit OpenAI’s GPT-5.6 Release for Government Review

The Trump administration has requested that OpenAI restrict the initial rollout of its advanced AI model, GPT-5.6, to a select group of government-approved partners. This move, reported by The Information and Axios, comes as federal officials aim to evaluate the model’s capabilities.

This marks the second instance this month where the U.S. government has intervened to control the release of a cutting-edge AI system. Previously, the administration ordered Anthropic to halt public access to its Claude Fable 5 and Mythos 5 models, citing national security concerns.

According to the reports, the White House’s Office of the National Cyber Director and the Office of Science and Technology Policy are urging OpenAI to limit GPT-5.6’s deployment. The administration is currently developing a framework for assessing advanced AI models before they are made widely available. Sources close to the discussions indicated that the request was primarily driven by GPT-5.6’s “Mythos-like” capabilities, rather than a broader change in the government’s AI policy.

This request follows President Trump’s executive order earlier this month, which directed federal agencies to establish a voluntary testing framework for advanced AI systems prior to their release. This directive emerged after weeks of internal deliberations regarding the structure of such a program.

The administration’s intervention also signifies a shift in the dynamic between leading AI developers and Washington. For years, AI companies have advocated for government regulation within the industry.

In 2023, OpenAI CEO Sam Altman testified before the Senate, urging lawmakers to establish a dedicated regulatory agency for advanced AI systems, arguing that independent oversight would eventually be essential. More recently, Anthropic CEO Dario Amodei echoed similar sentiments, contending that the most powerful AI models should undergo rigorous, government-backed evaluations before deployment due to their potential to facilitate sophisticated cyberattacks, aid in biological weapons research, and pose other national security threats.

These concerns have become increasingly formalized, with Anthropic, OpenAI, and Google each publishing proposals outlining potential governance structures for frontier AI. While their approaches vary, all three companies advocate for structured evaluations of the most capable models, enhanced transparency in safety testing, independent reviews of high-risk systems, and a more significant governmental role in overseeing AI development.

The administration’s intervention may also serve as a test case for the feasibility of applying the governance frameworks championed by major AI companies uniformly across the industry. Critics have raised concerns that if the largest AI developers influence the creation of rules that are then enforced inconsistently, frontier AI regulation could devolve into a form of regulatory capture, benefiting a select few companies and hindering competition.