The U.S. government has asked OpenAI to delay a broad release of its next frontier AI model, GPT-5.6, and instead implement a staggered preview process that would grant access on a customer-by-customer basis, according to reports from The Information and Reuters. The request, which OpenAI CEO Sam Altman communicated to staff, comes as the Trump administration cites security concerns over the uncapped deployment of highly capable AI systems. Access during the preview period would initially be limited to a vetted group of partners, with government agencies such as the Office of Science and Technology Policy and the Office of the National Cyber Director reviewing additional users.
The move marks a significant shift from voluntary safety testing to a more formalized, government-influenced access-control model. While OpenAI has not yet publicly announced GPT-5.6, its latest public release, GPT-5.5, has already undergone extensive safeguards and red-team testing. The intervention underscores a growing federal focus on mitigating national-security risks linked to advanced AI, particularly in areas like cyber operations, disinformation, and dual-use knowledge dissemination.
Separately, OpenAI is weighing a delay of its much-anticipated IPO from late 2026 to 2027, as reported by the New York Times. Internal discussions have leaned toward postponing the listing amid choppy market conditions and a tepid reception to SpaceX’s recent public debut. SpaceX shares tumbled from a peak of $225 to $153 after raising over $85 billion in its June IPO, dampening enthusiasm for high-valuation tech offerings. Advisers have warned that retail investor appetite may be insufficient to support a mega-scale debut.
CEO Sam Altman is pushing for a $1 trillion valuation at debut, a figure he has reportedly called a “nonstarter” to negotiate downwards. OpenAI’s last private round valued the company between $730 billion and $852 billion. Despite generating $13 billion in revenue last year and hitting a monthly run rate of $2 billion, the company posted a staggering net loss of $38.5 billion, largely due to $34 billion in compute and research expenses. With projected spending of $600 billion on hardware through 2030, the firm is exploring new revenue streams such as ads in ChatGPT and e-commerce partnerships with Shopify and Stripe, while retreating from unprofitable products like the Sora video app.
A crowded IPO pipeline adds further pressure. Rival Anthropic, valued at $965 billion, filed confidentially on June 1, and other companies like Strava, Discord, Kraken, and Oura have also filed, competing for investor attention. OpenAI’s Chief Financial Officer Sarah Friar had earlier raised concerns about the company’s financial footing, and the internal hesitation seemingly predates the confidential SEC filing.
For the broader tech and AI sectors, the dual developments signal that frontier model releases are becoming a matter of national policy, while the path to public markets for even the most prominent AI companies remains uncertain. The regulatory precedent could reshape the AI market, potentially creating a tiered system where only approved entities access the most capable models, a scenario that may influence institutional investment and startup dynamics across technology and adjacent markets.
