Donald Trump’s administration has asked OpenAI to restrict the initial release of GPT-5.6 to a group of clients approved by the White House, according to the agency Axios. This is the first time that the United States government has preemptively requested that a company limit access to an artificial intelligence model before its release to the public.
OpenAI agreed to a staggered launch, but said it does not intend to adopt this format permanently. The expectation is that access will be expanded after a few weeks of testing.
The request was made by the Office of the National Cybersecurity Director and the Office of Science and Technology Policy, which want to evaluate the model before wider distribution.
In an internal memo revealed by The InformationOpenAI CEO Sam Altman stated that the company collaborated with the government, but advocated for a more stable solution for future releases.
According to AxiosGPT-5.6 was considered by the American government to be comparable to Anthropic’s Mythos in cybersecurity capabilities.
In June, the government had already restricted the international distribution of Anthropic’s most advanced models for national security reasons. The concern is to prevent high-capacity systems from being used by criminals, foreign governments or malicious agents.
This month, Trump signed an executive order encouraging companies to voluntarily submit their systems to security assessments before public release, but the mechanism is still under construction.
