Who Gets to Decide
The White House has requested that OpenAI limit the release of its upcoming GPT 5.6 model to a small number of government-approved partners because of its advanced capabilities, a move that puts the machinery of state approval directly between the company and the public. OpenAI agreed to limit the model’s release as a path toward launching it publicly during what one source described as a “strange moment” with no true federal regulatory framework in place for new AI models.
That is the basic arrangement now: access to a new model is being filtered through government approval, while the rules for that approval remain unwritten. The result is a system where power moves by request, memo, and opaque coordination rather than any transparent public process.
The request came after the administration placed an export control order on Anthropic, leading the AI company to pull its latest most advanced models, Mythos and Fable. Those models raised fears in Washington and on Wall Street over their advanced cybersecurity capabilities, which some worry could lead to unprecedented safety risks. The people making and moving these systems are not the ones expected to absorb the consequences; that burden sits elsewhere, with ordinary people left to live under whatever “safety” the top layers decide to impose.
Customer by Customer
OpenAI and the administration view OpenAI’s latest model as “on par” with Mythos, according to the source. The Information, which first reported the Trump administration’s request, cited a memo OpenAI CEO Sam Altman sent to the company on Thursday, in which he said the government is approving access “customer by customer.”
“We’ve made clear to the U.S. government that this is not our preferred long term model, and will work with them and others in industry to achieve a more sustainable approach for future releases,” Altman said in the memo, according to The Information.
That phrase, “customer by customer,” captures the hierarchy cleanly: access is not being opened in a common public sphere, but parceled out through a gatekeeping process that runs upward through the state and outward through industry. The company says it is not its “preferred long term model,” but for now it is the model in force.
A White House official said the administration continues “to collaborate with frontier AI labs to develop shared approaches for addressing the challenges of scaling this technology.” OpenAI declined to comment. The language of collaboration does not change the structure underneath it: the White House is still the one asking, approving, and setting the terms.
What They Call Oversight
Though President Donald Trump signed an executive order earlier this month that asked AI companies with advanced models to voluntarily submit them for government review 30 days before release, the framework for that has not been established. In the interim, there is confusion among AI companies over who or which agency is directing AI regulation. The request to OpenAI came from the White House, while the export control ban on Anthropic came from the Commerce Department.
That split alone shows how little clarity exists in the apparatus now claiming to manage this technology. One arm of the federal machine issues a request, another issues a ban, and companies are left to interpret which authority is speaking for the whole. The public is not offered a stable rule; it gets a moving target.
Experts said the government should be involved in conversations about AI safety, especially those that impact national security, but that there is no transparent, consistent framework for regulating AI and that could stifle the industry. Brad Carson, head of Public First, a bipartisan pro-AI safety super PAC, said last week, “The Fable episode shows the need for clear regulations. Right now, you have an ad hoc, personalized, opaque, possibly lawless approach. It is certainly appropriate for the government to recall dangerous products, including AI models, but it has to be done in a way consistent with transparency and basic fairness.”
That complaint lands on the same problem from another angle: the current setup is ad hoc, personalized, and opaque. Whether framed as safety, national security, or industry management, the process remains concentrated in the hands of institutions that decide who gets access, who gets blocked, and who gets to call it order.