OpenAI confirmed in a post earlier this week that it agreed to a White House request to release its next model, GPT-5.6, only to a small group of government-approved partners, citing the model’s advanced capabilities. The request came from the White House’s Office of the National Cyber Director and Office of Science and Technology Policy, according to a source familiar with the matter, and marks the first time the U.S. government has preemptively asked an American AI company to hold back a model before launch.

The move puts a major business decision in the hands of Washington at a moment when the rules for advanced AI are still being written.

The request followed talks between OpenAI chief executive Sam Altman and Commerce Secretary Howard Lutnick, who wanted to be sure the relevant parts of the government had tested and approved the model. A source told Axios the government stepped in because GPT-5.6 has “Mythos-like” capability — a reference to the advanced systems from rival Anthropic.

That comparison matters because of what happened to Anthropic just days earlier. The administration placed an export control order on the company, which led it to pull its most advanced models, Mythos and Fable. Those models had stirred fears in Washington and on Wall Street over their cybersecurity abilities and the safety risks that could follow.

OpenAI made clear it does not want this to become the norm. In its Friday post, the company said this kind of government access process should not become the long-term default, warning it keeps the best tools from users, developers, businesses, cyber defenders and global partners who need them. In an internal memo first reported by The Information, Altman said the government is approving access customer by customer, and that the company would push for a more sustainable approach for future releases. OpenAI said it hopes to make the model widely available in the coming weeks.

For the AI business, the bigger problem is confusion over who is actually in charge. The request to OpenAI came from the White House, while the export control order on Anthropic came from the Commerce Department. President Donald Trump signed an executive order earlier in June asking AI companies with advanced models to voluntarily submit them for government review 30 days before release, but the framework for that review has not been built. In the meantime, companies are unsure which agency is directing AI regulation.

That uncertainty carries a real cost for the industry. Companies pour billions into building and launching these models, and a launch that can be paused or narrowed at the government’s request changes the math for investors, enterprise customers and developers who plan their own products around release dates. A model that only a handful of approved users can touch generates far less revenue than one sold broadly.

The policy reversal is striking. The administration started out hands-off on AI, scrapping Biden-era rules that required safety reviews of frontier models. It has since shifted hard the other way — clashing with Anthropic, blocking foreign nationals from its most advanced systems, and imposing nominally voluntary reviews.

Safety advocates say government involvement is appropriate but the process needs to be transparent. Brad Carson, who leads the pro-AI-safety group Public First, said the Fable episode shows the need for clear rules and warned that the current approach is ad hoc and opaque. He said it is fair for the government to recall a dangerous product, including an AI model, but it has to be done with basic fairness.

For now, businesses building on these tools are left waiting. OpenAI says it will keep working with the administration to find a steadier path for future launches. Until there is a clear referee, every major model release becomes a negotiation with Washington — and that is a new kind of risk for one of the fastest-growing industries in the country.

JBizNews Desk | New York

© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.