The Trump administration has taken an unusual step and directed artificial intelligence (AI) leader OpenAI to restrict the launch of its next-generation model, GPT-5.6 – the first time the federal government has preemptively stepped in to limit release of a commercial American AI model because of national security and cybersecurity anxieties.
According to internal company memos first reported by The Information and confirmed by subsequent intelligence, OpenAI CEO Sam Altman informed staff during a recent Q&A session that GPT-5.6 will not see a traditional public deployment. Instead, the model will transition into a highly restricted preview, with the federal government approving access on a “customer-by-customer” basis.
The mandate originated from discussions with the White House Office of the National Cyber Director and the Office of Science and Technology Policy. Federal officials grew alarmed after briefings and early previews revealed that GPT-5.6 possesses advanced capabilities “on par” with Anthropic’s highly scrutinized Mythos and Fable models.
Security experts warn that so-called “frontier cyber tools” possess the unprecedented ability to autonomously identify and exploit complex software vulnerabilities at speeds human analysts cannot match. This raises critical fears that the technology could be weaponized by nation-state spies, cybercriminals, or rogue insiders to launch automated ransomware attacks.
The intervention follows a disruptive sequence of events for the AI sector. Several weeks ago, a rare Commerce Department export control directive forced competitor Anthropic to revoke access to its most advanced models, Claude Mythos and Fable 5. Behind the scenes, Altman met with Commerce Secretary Howard Lutnick on Wednesday to ensure relevant federal entities could thoroughly vet OpenAI’s safeguards before any wider distribution is authorized.
The administration’s aggressive maneuvers highlight a growing tension between Silicon Valley’s rapid development cycle and Washington’s lagging regulatory infrastructure. As President Donald Trump signed an executive order earlier this month asking AI firms to voluntarily submit advanced models for a 30-day pre-release review, the official testing protocols remain stalled by political infighting.
The current ad-hoc enforcement has sparked deep confusion across the tech sector. While the commerce department leveled bans against Anthropic, the White House directly pressured OpenAI, exposing a fractured regulatory landscape.
“Right now, you have an ad hoc, personalized, opaque, possibly lawless approach,” said Brad Carson, head of the bipartisan pro-AI safety super PAC Public First. “It is certainly appropriate for the government to recall dangerous products, but it has to be done in a way consistent with transparency and basic fairness.”
In his staff memo, Altman signaled compliance but expressed clear dissatisfaction with the ongoing regulatory bottleneck, stating that OpenAI has made it clear to the U.S. government that individual client screening is “not our preferred long-term model.”
OpenAI reportedly hopes to secure permission for a broader public release within a few weeks if the initial partner trials go smoothly. However, the company remains caught in a high-stakes bind, striving to navigate opaque domestic security demands while racing to outpace heavily subsidized, open-source AI rivals in China.
While OpenAI faces an increasing regulatory landscape in Washington amid market volatility, the company is leaning toward delaying its highly anticipated initial public offering until 2027, The New York Times reported Thursday. The decision marks a shift for Altman, who previously favored a quicker timeline, and aligns with Chief Financial Officer Sarah Friar’s long-standing preference for a later listing.
The AI pioneer completed its necessary corporate restructuring last October, transforming into OpenAI Group PBC. While initially weighing a late 2026 debut at a valuation up to $1 trillion — following an $852 billion valuation in March — the company is now practicing caution.
Internal deliberations cite heavy cash burn, massive compute commitments, and rigid public reporting requirements as key hurdles. Furthermore, recent market turbulence has spooked executives; SpaceX’s recent $75 billion IPO saw its early gains swiftly erased, signaling potentially weak demand from retail investors.
An OpenAI spokesperson stated that an IPO is not their current focus.