The Trump administration has asked OpenAI to delay the broad release of GPT-5.6 due to safety concerns. The company, led by Sam Altman, will initially provide the model to a limited number of clients, according to reports from Reuters and The Information.
Sources indicate that the request came from the White House's National Cyber Director's Office and the Office of Science and Technology Policy. These agencies have asked to restrict the launch of GPT-5.6 while the administration develops a framework for testing and assessing the safety of new AI models.
OpenAI has not publicly commented on the limited release, but The Information reports that Altman informed employees about it. A source familiar with the matter confirmed to Reuters the government's request for a staggered release of GPT-5.6.
Changes to the Release
According to The Verge, this will involve a small group of corporate clients. The federal government will determine which customers can access the model during the preview phase.
"We have made it clear to the US government that this is not our preferred long-term model, and we will work with them and other industry stakeholders to develop a more sustainable approach to future releases," Altman said in a message to employees, as reported by The Information.
Axios cites a source stating that OpenAI had been collaborating with the administration on the GPT-5.6 release. The White House was reportedly aware of the new model's capabilities and had the opportunity to evaluate them in advance. The report also noted that on June 25, Altman discussed this matter with US Secretary of Commerce Howard Latnik, who wanted to ensure that relevant government entities had tested and approved the model.
Why the Government Intervened
Media sources linked the request to the new model's capabilities in sensitive areas. According to CNN, OpenAI and the administration view GPT-5.6 as comparable to Mythos. An unnamed expert cited by Axios echoed this comparison, emphasizing that authorities want to ensure adequate protective measures are in place for tools of this class.
A White House representative told CNN that the administration continues to collaborate with developers of advanced AI models to establish common approaches to scaling the technology.
The request to OpenAI coincided with a broader restructuring of US policy regarding advanced artificial intelligence. On June 2, Trump signed an executive order promoting advanced artificial intelligence innovation and security. The document does not mandate licensing for AI model releases but directs several agencies to develop a classified benchmarking process to evaluate their advanced cyber capabilities.
The order also includes a voluntary mechanism for engaging with developers, allowing companies to provide the government with advanced models up to 30 days before release. For trusted organizations, this access must be accompanied by confidentiality, cybersecurity, insider risk protection, and intellectual property safeguards.
Contrast with Anthropic
The situation with OpenAI appears less severe than the recent case involving Anthropic. On June 9, the startup released two versions of its Claude model family. The company described Fable 5 as a Mythos-class solution that is safe for general use, while Claude Mythos 5 is a "private" base model with relaxed restrictions in certain areas.
However, on June 12, Anthropic disabled the models due to a US government directive under export control. The company stated that the document formally prohibits access to any foreign nationals, including foreign employees. Later, media reports linked the restrictions to China, and some sources indicated that another trigger was reports of circumvention of Fable 5's protections.
In light of these decisions, market participants and experts pointed to the lack of a transparent procedure. CNN quoted Public First CEO Brad Carson, who described the situation surrounding Fable as an example of the need for clear rules.
"The Fable episode highlights the need for clear regulation. Right now, you have a special, personalized, opaque, and possibly illegal approach," he stated.
Carson added that the government may intervene in the case of dangerous products, including AI models, but such processes must adhere to principles of basic fairness.
OpenAI's Position on Regulation
In early June, OpenAI presented a framework for governing advanced AI. The company proposed strengthening the Center for AI Standards and Innovation and developing a broader plan for the government's resilience to the risks posed by next-generation models.
"The federal government must now build on this foundation a sustainable federal framework capable of evolving alongside the technology itself," the document states.
In a separate document, OpenAI outlined its approach to assessing and mitigating risks in areas such as cyber incursions, CBRN threats, harmful manipulation, loss of control, model accountability, security risk management, incident response, and external expert evaluation.
Previously, the company confidentially filed an S-1 form with the US Securities and Exchange Commission for a potential IPO. According to Reuters, OpenAI is preparing for a public stock offering with a valuation of $1 trillion.
The Wall Street Journal, citing sources, reported on the possibility of the company going public in September and mentioned that Goldman Sachs and Morgan Stanley are involved in preparing the deal. In a new statement, OpenAI did not confirm these timelines. According to NYT, advisors suggested two scenarios to the management: wait until 2027 for a $1 trillion valuation or agree to an earlier listing with a lower threshold. Altman reportedly rejected the latter option.
On June 22, OpenAI launched the full version of GPT-5.5-Cyber, a specialized model for finding, verifying, and fixing vulnerabilities. The release occurred amid restrictions surrounding Anthropic.