Limit OpenAI’s latest model launch

Trump administration steps in to limit OpenAI’s latest model launch

OpenAI will now make its soon-to-be-released GPT-5.6 model available to a small group of government-approved partners.

By Cheyenne Haslett

OpenAI’s newest artificial intelligence model will be available only to a small group of U.S. companies and organizations approved by the Trump administration, two people familiar with the discussions said, further signaling the Trump administration is tightening the reins on AI regulation.

Details about the latest model, GPT-5.6, have not yet been released publicly. The company did not initially plan to restrict the release of the general-use model, but changed course at the White House’s request and in consultation with the Office of Science and Technology Policy and the Office of the National Cyber Director, the two people said, granted anonymity to describe internal discussions about OpenAI’s latest product release.

The Information first reported on the administration’s intervention in GPT-5.6’s launch, citing a leaked memo from OpenAI CEO Sam Altman to staff on Thursday, in which he said that while the controlled release was not the “preferred” method for announcing its latest AI model, the company was hopeful it could work with the government to develop a more “sustainable” approach to unveiling its iterations of models going forward.

The White House’s rigorous oversight in the rollout of new AI models comes as it continues to spar with OpenAI competitor Anthropic over its newest model, Fable 5. Anthropic was forced to pull Fable 5 just days after its public release, following concerns that the model’s guardrails were less secure than the company had suggested.

The White House made its first significant step in regulating AI earlier this month, when President Donald Trump signed an executive order to manage the release of powerful new AI models with advanced hacking capabilities. The order was sparked by Anthropic’s most advanced model, Claude Mythos, which was made available only to a select group of trusted researchers and cybersecurity companies, as the company warned that the model could be used to launch major cyberattacks if it fell into the wrong hands.

OpenAI has similarly released its most cyber-capable model, GPT-5.5-Cyber, to a small number of companies and organizations.

But the executive order, meant to create a process for evaluating the most advanced models, is still in its nascent stages and hasn’t yet established clear rules of the road for companies as they continue to race ahead with the latest AI advancements, competing with each other and with adversarial countries like China.

The result is an open-ended and confusing regulatory landscape for AI companies to navigate, as the administration — which previously pledged to prioritize innovation over regulation — now faces down the potential risks posed by rapidly developing AI.

In a statement, the White House said it “continues to collaborate with frontier AI labs to develop shared approaches for addressing the challenges of scaling this technology.”

OpenAI did not immediately respond to a request for details on the rollout of GPT-5.6.

The June executive order calls on companies to voluntarily submit models to the government for review 30 days before their planned release. The vetting period is meant to give federal agencies time to gauge what threats the products may pose to sensitive financial, national security and other critical systems.

But the details of how models will be evaluated and which models need to be reviewed are still being worked out in White House meetings with the companies over the next several weeks, as directed by the executive order.

Some have criticized the administration’s approach and called for clearer guidelines.

“AI is licensed now, but the requirements change constantly and are always a secret, even to the administration itself, which will discover the rules spontaneously in real time as it reacts to events,” Dean Ball, a former Trump administration AI policy adviser who recently joined OpenAI, wrote on X in mid-June.

A week later, as Fable 5 remained offline but before OpenAI faced its own showdown with the government, Ball had a prescient prediction: “I’d assume the whole AI industry in America is effectively frozen from new public releases until USG resolves the Fable situation they have stumbled into,” he wrote.