The Trump administration has asked OpenAI to delay the release of its next-generation artificial intelligence model and grant early access only to a select few approved by the government, following a similar move involving Anthropic's Mythos. It marks the first case of regulation imposed before a new model has even launched, signaling that the U.S. government's oversight of powerful AI models is growing tighter.
OpenAI's Next-Generation Model Regulated Before Launch
Axios reported Tuesday, citing sources familiar with the matter, that the White House Office of the National Cyber Director (ONCD) and the Office of Science and Technology Policy (OSTP) asked OpenAI to limit the release scope of its next-generation model, "GPT-5.6," while the administration builds a framework to test and evaluate the security of new AI models.
OpenAI Chief Executive Officer Sam Altman reportedly met with U.S. Commerce Secretary Howard Lutnick on Monday to discuss GPT-5.6. Sources said Lutnick requested time to verify that all relevant agencies within the government had tested and approved the model.
This is the first case in which a release restriction was requested preemptively, before a new AI model has been formally launched. The Trump administration had previously ordered access blocked for foreign nationals at home and abroad immediately after Anthropic's Mythos 5 and Fable 5 launches, effectively imposing an export ban. GPT-5.6 is known to have performance comparable to Mythos.
OpenAI expressed dissatisfaction with the delayed release and stated that it wants to make the model available to the public within weeks. According to an internal memo obtained by U.S. tech outlet The Information on Tuesday, OpenAI CEO Sam Altman said, "We have made clear to the U.S. government that this is not our preferred approach over the long term," adding, "For future model releases, we will work with the government and other companies in the industry to achieve a more sustainable approach." He also noted, "We hope to be able to fully release GPT-5.6 to the public a few weeks after the early release to our government partners."
As Chinese AI Surges, Launching New Models Grows Harder
Axios analyzed that U.S. AI labs are caught in a dilemma. They are under pressure to rush out new models to compete not only with other U.S. AI models but also with Chinese open-source models whose performance is becoming more advanced by the day. At the same time, concerns are growing among security officials and corporate leaders about the fallout if malicious actors—such as spies, cybercriminals, or insider traitors—were to get their hands on these high-performance models.
Earlier this month, President Trump signed an "AI security executive order" directing several government agencies to establish voluntary testing procedures that AI companies must go through before releasing new models. A source said, "All of this (release regulation, etc.) is happening to Mythos-class models," adding, "Because the models are so powerful, the administration wants to make sure companies have put appropriate safeguards in place."
