The US government has asked OpenAI to delay the broad release of its upcoming AI model, GPT-5.6, citing national security concerns, according to The Information. According to the proposed plan, the model would be released in phases, with the US government approving early users on a case-by-case basis.
The report said OpenAI CEO Sam Altman informed employees that GPT-5.6 would first be made available to a limited group of partners rather than launching widely. During this preview period, access will be granted on a customer-by-customer basis with government approval.
According to The Information, the request came from the White House’s Office of the National Cyber Director and the Office of Science and Technology Policy, as the US government increases its scrutiny of frontier AI models and their potential security implications.
AIM reached out to OpenAI for a comment, but hasn’t received a response at the time of publication.
The arrangement gives the US government an unprecedented role in determining early access to a frontier AI model before its wider launch. While the Donald Trump administration has largely promoted rapid AI development, it has also introduced measures to review the capabilities of the most advanced systems before they become widely available.
The report added that OpenAI believes the limited preview is the quickest path toward an eventual broader launch. By working with the government during the initial deployment, the company hopes to address security concerns before expanding access.
The development follows heightened scrutiny over the risks posed by increasingly capable AI systems, including concerns about cybersecurity, misuse by hostile actors, and the possibility that advanced models could expose sensitive capabilities if released without adequate safeguards.
Recently, the Trump administration took a tougher stance toward Anthropic’s frontier models. It recently issued an export-control directive banning all foreign nationals worldwide from accessing Dario Amodei-led company's most advanced AI models, Fable 5 and Mythos 5, citing national security concerns.
OpenAI is also considering postponing its initial public offering (IPO) until 2027 as executives debate whether current market conditions justify going public before achieving a higher valuation, according to a report by The New York Times.
The ChatGPT maker confidentially filed paperwork for a US IPO earlier this month and initially targeted a public listing later this year. However, advisers have reportedly urged the company to wait as technology stocks remain volatile and investor appetite for large offerings has weakened.
OpenAI is seeking a valuation of as much as $1 trillion, and CEO Sam Altman has resisted suggestions to lower that target in exchange for a faster public debut. The company is now reportedly weighing whether to wait until market conditions improve rather than accept a lower valuation.
Your reaction