NEW YORK — The U.S. government effectively put a leash on OpenAI’s newest AI family before it could run free. The GPT-5.6 series, revealed today, launches in a restricted “limited preview.” That is not what OpenAI planned. CEO Sam Altman said the company was aiming for open access.
The government asked otherwise. Three models make up the series.
Sol sits at the top. OpenAI calls it smart, efficient, and a significant step forward. Its price matches GPT-5.5 exactly: unchanged.
Terra sits in the middle. It delivers performance comparable to GPT-5.5 at half the cost.
Luna, the hyper-fast model, had no pricing in the release. The tiers are distinct. The restriction applies to all of them.
The government’s request follows national security scrutiny. The Trump administration previously used export controls to take rival models offline. Later, those models were allowed with conditions.
Now OpenAI is coordinating with Washington ahead of an upcoming White House cyber Executive Order. Altman did not specify which agency made the request or how long the preview would last.
He said the company is working to get to general availability as fast as possible. Altman framed the limited preview as consistent with company strategy. “I think it is quite reasonable to roll out models—especially as they reach significant new levels of capability—in this way,” he said.
He described it as fitting with OpenAI’s “long-held strategy of iterative deployment.” But he added: “this isn” — the quote cuts off in the release. The implication is clear.
This was not their plan. So the core fact of the day is not that Sol is smarter. It is not that Terra costs less.
The core fact is that the U.S. government stepped in before launch. That is new. OpenAI has faced regulatory pressure before, but this is a direct, preemptive request to hold back a product.
The company complied. No public fight.
No leaked internal memos. Just Altman delivering the news in a single sentence: “at the request of the US government, it is launching today in limited preview instead of the open access launch we were planning on.” That sentence changes the story. This is no longer a product launch.
It is a product preview, with government sign-off pending. The models exist.
They work. But you cannot simply sign up and use them. Not yet.
OpenAI did not say how many users are in the limited preview. It did not say which industries or researchers got early access. It did not say what conditions the government set.
The release is thin on those details. What is clear: the company accepted the restriction publicly and without visible pushback.
Altman called it reasonable. This is a shift. Previous OpenAI launches — GPT-3, GPT-4, GPT-5.5 — rolled out to wide audiences.
Some had safety delays. Some had usage caps.
None had a government-requested limited preview announced on day one. This is the first. The Trump administration’s earlier actions against rival models set a precedent.
Those models were rendered offline via export controls. Later they were allowed with conditions. Now OpenAI is working alongside Washington, not against it.
The upcoming White House cyber Executive Order appears to be the context. Altman did not elaborate on what the order might require.
For now, the GPT-5.6 series is real but restricted. Sol is smart, efficient, and costs the same. Terra matches GPT-5.5 performance at half the price.
Luna exists. And the U.S. government has its hand on the release valve.
That is the story.
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “U.S. Government Restricts OpenAI’s GPT-5.6 Series to Limited Preview”, “description”: “NEW YORK — The U.S. government effectively put a leash on OpenAI’s newest AI family before it could run free. The GPT-5.6 series, revealed today,…”, “datePublished”: “2026-06-28T07:06:21+00:00T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Noah King”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}}
