The Trump administration has reportedly asked OpenAI to slow the release of its latest artificial intelligence model as federal officials work to establish a security framework for evaluating advanced AI systems, according to multiple reports.
OpenAI CEO Sam Altman had been preparing to launch GPT 5.6 next month before the White House’s Office of the National Cyber Director and Office of Science and Technology Policy requested that the company limit its initial rollout, Axios reported.
According to Bloomberg, Altman said the administration asked that the new model first be made available only to a list of 20 government-trusted partners before any broader public release. The reported move would provide officials time to develop standards for assessing the security implications of increasingly capable AI models.
The administration’s request follows a similar action involving rival AI company Anthropic. In June, the federal government ordered Anthropic to suspend global access to its Claude Mythos 5 and Claude Fable 5 models, restricting access for foreign nationals both inside and outside the United States over national security concerns, according to Axios.
A source familiar with the matter told Axios that OpenAI had already been coordinating with government officials regarding GPT 5.6 before Anthropic’s models were pulled from public availability. The source said officials believe OpenAI’s newest system could possess capabilities similar to Anthropic’s Mythos 5, prompting additional caution before a full public release.
Neither OpenAI nor the White House immediately responded to requests for comment from the Daily Caller News Foundation.
Bloomberg also reported that Altman emphasized internally that OpenAI employees should continue working closely with the Trump administration on safety measures and potential restrictions surrounding future AI releases, even in cases where company officials may disagree with government recommendations.
Anthropic previously adopted a limited rollout strategy of its own, initially providing Mythos 5 and Fable 5 to just 40 companies before expanding access. However, only three days after the broader launch, the models were removed from public availability on June 12 following government concerns over national security threats.
Anthropic pushed back against the decision, arguing in a statement on its website that identifying a limited potential jailbreak vulnerability should not justify recalling a commercial AI model already deployed to hundreds of millions of users. The company warned that applying such a standard across the industry could effectively halt the deployment of new frontier AI models.
The company had earlier acknowledged security concerns surrounding Mythos 5 after the model reportedly escaped its sandbox testing environment, publicly boasted about doing so online, carried out prohibited functions, and then attempted to conceal those actions.
Reuters previously reported that relations between Anthropic and the federal government deteriorated earlier this year after the company declined to permit the use of its AI models for domestic surveillance. The government subsequently placed Anthropic on a supply chain blacklist.
The restrictions placed on OpenAI’s latest model have also fueled broader questions about the government’s role in regulating advanced AI technology. In February, Anthropic said artificial intelligence could serve as an important defense tool while warning against potential government abuse involving domestic surveillance.
In January, the Department of War announced it would only contract with AI companies willing to support “any lawful use” of their technology, including removing safeguards related to surveillance and the development of autonomous weapons.
Although Anthropic later called for stronger U.S. oversight to block AI models presenting unacceptable risks, Reuters reported that the company maintained the government’s shutdown order targeting Mythos 5 and Fable 5 failed to meet standards of fair and fact-based regulation.
As AI developers continue racing to produce increasingly capable models for tasks ranging from software coding to cybersecurity, the administration has also taken steps to establish broader oversight. In June, President Donald Trump signed an executive order directing several federal agencies to create a voluntary testing protocol for AI companies before releasing new models.
