The US government just told OpenAI who’s allowed to use the next GPT 5.6 model

The release of leading AI foundation model technologies has reached a watershed moment. Just two weeks after Anthropic received a directive from the US government to suspend access to Fable 5 and Mythos 5, the White House this week has mandated that OpenAI place limitations on the release of its upcoming GPT 5.6 model due to cybersecurity concerns.

This new directive stipulates that OpenAI’s GPT 5.6 can be made available only to a limited number of government-approved partners due to the advanced nature of its capabilities, according to a CNN source familiar with the situation.

Initially reported on The Information Wednesday, the report references a staff Q&A session hosted by OpenAI CEO Sam Altman. A post on X stemming from the report explained that the federal government would be “approving access customer by customer during this preview period” for GPT 5.6. 

Closing the door stokes & feeds the fire

Collin Hogue-Spears, senior director and distinguished technology expert at Black Duck, tells The New Stack that gating the closed API does not reduce capability; it shifts demand to locations where the U.S. government lacks jurisdiction.

“Z.ai’s GLM-5.2 already beats GPT-5.5 on several coding benchmarks and runs on hardware that developers themselves control, so access controls on the closed model push serious teams toward an alternative they can host themselves,” Hogue-Spears says. “Every customer-by-customer approval makes that alternative more attractive, not less, so the security goal and the market reaction now run against each other.”

“…the security goal and the market reaction now run against each other.”

Hogue-Spears advises that the absence of federal regulation is the “core problem for builders,” who need a fixed rule to plan against rather than discretionary action applied on a model-by-model basis. At the same time, he says teams are navigating a patchwork of state AI laws with conflicting definitions and deadlines, which, in his view, is a “result of the lack of leadership” from Washington.

Teams are navigating a patchwork of state AI laws

While it’s true that no formalized federal regulatory framework exists to govern and oversee the ongoing release of new AI models, President Donald Trump signed an executive order on June 2 asking AI companies to submit new model releases 30 days before release.

According to the Federal Register, the official daily journal of the US government, Executive Order 14409 of June 2, 2026, comes because advanced AI capabilities make the nation stronger.

“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.” —Sam Altman.

“But [they] also introduce new national security considerations that require coordinated action across executive departments and agencies (agencies) and components. As these capabilities evolve, my Administration will continue to work closely with industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to our country,” details the executive order signed by Trump.

In a snippet from the OpenAI CEO’s memo seen by The Information, Altman is reported to have been less than happy: “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.”

What software developers think of government directives

The software engineering community has, unsurprisingly, been unimpressed. CIO at Interconnected Capital and ex-GitHub and Obama White House/Commerce Dept executive Kevin S. Xu posted on X saying, “The government will be approving access to GPT-5.6 customer by customer? [That makes China’s model pre-release registration scheme look almost laissez-faire.”

Echoing that sentiment, Charming-Author4877 writes on Reddit, “By the time they release GPT 5.6 we’ll hopefully have the next GLM, Qwen, Deepseek or Kimi that beats it. [The] US is within months of losing the lead in AI.”

Redditor Punter1965 agrees with the general sentiments expressed thus far. He has suggested that all this will do is drive companies to choose open-source or Chinese models rather than deal with these headaches. 

“Basically, Trump’s spat with Anthropic and its spillover will only diminish America’s AI lead,” they noted. “Unfortunately, by the time this gets worked out (either by bribery or at the Supreme Court of the United States), the damage may be done, and we could find China ahead of us.

With developers joking that next, we’ll see the “government is approving GPT-5.6 usage prompt-by-prompt”, the mood is not upbeat.

“My concern is the long-term impact: if access to cutting-edge AI increasingly depends on where you live or which organization you work for, we risk moving away from the open innovation model.” – Sylwia Laskowska

The Chrome-style release honeymoon period is over

Gdańsk, Poland-based senior software engineer Sylwia Laskowska tells The New Stack that until now, new AI models have been released much like new versions of Chrome or React, i.e developers could start experimenting with them almost immediately. 

“If access to cutting-edge AI increasingly depends on where you live or which organization you work for, we risk moving away from the open innovation model.”

“If frontier AI models are now becoming strategic assets, similar to advanced semiconductors, that’s a fundamental shift,” Laskowska says. “For the average developer, this may not make much difference at first. We’ll probably just receive more stable, thoroughly tested releases. My concern is the long-term impact: if access to cutting-edge AI increasingly depends on where you live or which organization you work for, we risk moving away from the open innovation model.”

CISO at Sumo Logic, Jeremy Powell, tells The New Stack that the real story here is about capability, rather than politics. He thinks that, from a security perspective, when a lab and the government both decide a model is too powerful to release widely, that tells us the offensive bar just moved. 

“Restricting access to frontier AI models slows the spread — it doesn’t undo the capability.”

“For developers, that ability to connect up different issues between components to achieve an exploit would mean that they have to update those components more rapidly,” Powell says. “Restricting access to frontier AI models slows the spread — it doesn’t undo the capability. At the same time, that improvement in performance will make a difference to developers around how they build software and create new agents that can handle more complex tasks.”