In Washington, the Department of Commerce has reportedly cleared OpenAI to expand access to its GPT-5.6 model family, ending a tightly government-vetted preview that had limited the models to a small group of partners. If the broader rollout happens, it would be the latest sign that federal agencies are directly shaping when and how the most powerful commercial AI systems reach customers.
Commerce Signs Off After CAISI Tests
The Commerce Department reportedly gave OpenAI the go-ahead for a wider launch after additional testing and meetings in Washington, and OpenAI "expects to do a wide release" this week, according to Axios. The testing was reportedly performed by the Department's Center for AI Standards and Innovation, which has been designated to run pre-deployment technical reviews of frontier models, per the Center for AI Standards and Innovation.
OpenAI Calls Limited Preview A Short-Term Move
As detailed in a GPT-5.6 preview system card from OpenAI, the company launched three models, Sol, Terra and Luna, in a limited preview while coordinating with federal reviewers. OpenAI said it plans to make the models "generally available in the coming weeks" and has pushed back on the idea that government-curated access should become the norm.
Anthropic Run-In Set The Template
The administration's more hands-on posture followed a mid-June export-control order that forced Anthropic to suspend access to its Fable 5 and Mythos 5 models after researchers demonstrated a technique that bypassed some safeguards. Anthropic and the Commerce Department later announced that the controls were lifted once the company implemented new safeguards and the government had reviewed them, per reporting in The Washington Post.
Why Washington Is Locked In
All of this is playing out under the White House's June 2 executive order, which created a voluntary framework for early government access to "covered frontier models" and asked agencies to build repeatable benchmarking and review processes. The text of the order helps explain why labs, Commerce and CAISI have been negotiating release timing and customer lists before wider public launches, per the White House.
What Comes Next
OpenAI's system card reiterates that broader availability is planned in the coming weeks and outlines the safeguards the company used during the preview, so the next days should clarify how quickly the wider rollout will proceed. D.C. contractors, cyber teams and enterprises that rely on ChatGPT for code, analysis or infrastructure monitoring will be watching OpenAI's developer pages and Commerce guidance for exact timing and any remaining gating.
For Washington readers, the episode is a reminder that model releases are now policy events as much as product launches. The Commerce Department's offices and OpenAI's local presence will be the focal points as this government-mediated release process settles in as part of how frontier AI reaches the world.
