The White House Just Put A Lock On America’s Best AI

Trump’s White House is now deciding who gets early access to OpenAI’s newest model, and that has raised fresh alarms about federal control over private technology.

Quick Take

The government asked OpenAI to limit GPT-5.6 access to a small group of approved partners.^[1]^[4]
OpenAI told staff the government will approve access “customer by customer.”^[1]^[7]
The reporting ties the request to cybersecurity and advanced model capabilities.^[2]^[5]
The case adds to growing debate over whether Washington is creating a back-door licensing system.^[4]^[14]^[15]

Government Review Comes Before Broader Release

OpenAI is limiting GPT-5.6 rollout after talks with the Trump administration, according to multiple reports. The company said the model will first go to a small group of trusted partners, with government review shaping who gets in and when. That move follows a White House push for early review of frontier models and fits a broader trend toward tighter control of advanced artificial intelligence.^[4]^[6]^[14]

According to reporting cited by several outlets, Sam Altman told staff that the government is approving access “customer by customer.” The same reports say the White House asked for the staged release because the model’s advanced capabilities raised security concerns. That means the issue is not just speed of release. It is who gets to judge whether a private company’s product is safe enough for the public.^[1]^[2]^[7]

Why Officials Want a Slower Rollout

The government’s concern centers on the model’s ability to support high-end work in cyber defense and biology. One report says OpenAI’s internal safety material rated all three GPT-5.6 variants as “High” in both cybersecurity and biological capability. Another says the Trump administration had already ordered a 30-day government review window for advanced models, which helps explain why officials wanted a slower, more controlled launch.^[2]^[5]^[15]

That approach may sound like common sense to some readers who want bad actors kept away from powerful tools. But it also creates a major problem for trust and fairness. When the government can quietly steer access to a private model, there is no clear public standard, no open hearing, and no obvious limit on how far the gatekeeping can go.^[5]^[6]^[14]

What This Means for OpenAI and the Market

OpenAI has not said this restriction will become permanent, and the company reportedly told the government it does not want this to be the long-term model. Even so, the temporary limits show how much leverage Washington now has over frontier artificial intelligence. If the government can slow one release today, it can pressure the next one tomorrow. That is why critics see a creeping licensing regime rather than a narrow safety review.^[4]^[14]^[15]

Polymarket has ChatGPT’s 5.6 launch being pushed back to mid July after the government told OpenAI to delay the release. This is the second government interference in frontier model releases in the last month.

“To me this is more significant than the Fable ban. When the… pic.twitter.com/BI9AVn4zCS

— Counterparty (@counterpartytv) June 26, 2026

The fight also matters beyond OpenAI. Competitors, open-source builders, and investors are watching closely because a case-by-case approval system can shape who wins the next wave of AI. Supporters of open markets will see a warning sign in that power. If regulators can pick favored partners for the most advanced tools, then innovation may depend less on merit and more on political access.^[4]^[6]^[17]