According to The Information and Axios, the Trump Government has requested OpenAI to tighten the initial scope of GPT-5.6, open only to a few government-accredited partners. Federal officials are using this to assess the model and to advance a test framework for advanced AI systems.
The White House has requested a moratorium on the expansion.
The report mentions that this request comes from the Office of the Director of the White House National Network and the White House Office of Science and Technology Policy. The Departments would like to complete the GPT-5.6 assessment prior to a wider deployment.
This is the second time the United States Government has been involved in a front-line model released this month. Previously, the Government had requested Anthropic to suspend public visits to Claude Fable 5 and Mythos 5 for national security concerns.
This month's executive order. Response
This action followed the signing of an executive order by Trump earlier this month. The Executive Order requires the federal agencies to establish a voluntary test framework for the advanced AI system prior to its issuance.
According to the report, this Government request does not necessarily imply an overall shift in AI policy, more directly because officials consider GPT-5.6 to have a level of competence similar to “Mythos” and therefore wish to control access before deciding on subsequent arrangements.
Technology has long advocated more scrutiny.
Over the past few years, several head AI companies have been calling on the Government to strengthen the regulation of front-line models. The CEO of OpenAI, Sam Altman, advocated, at the Senate hearing in 2023, the establishment of a specialized regulatory body for the advanced AI system.
CEO Anthropic Dario Amodei has since also indicated that the most capable model should be subject to a more rigorous, government-supported assessment prior to deployment, as such systems may be used in high-risk scenarios such as complex cyberattacks and biological weapons research.
OpenAI, Anthropic and Google have also recently published the Frontline AI governance programme. Despite different paths, several companies generally supported structured assessments of high-capacity models, increased transparency in security testing and the introduction of more independent reviews.
Implementation measures or next focus
According to the report, this government intervention will also test whether these governance ideas can be consistently implemented within the industry. Critics are concerned that if the rules are largely driven by front-line companies, their implementation is uneven, they may make regulation more beneficial for a small number of large enterprises and raise the threshold for those who come later.