OpenAI received U.S. government approval for a broad GPT-5.6 rollout

In July 2026, the U.S. government gave permission for broad release of GPT-5.6—the most powerful model deployed by OpenAI. For several weeks, the model remained in closed preview for approximately 20 partners while Washington conducted state assessment within a new oversight regime for frontier AI systems.

Why was GPT-5.6 delayed at launch?

The delay is explained by the adoption in the U.S. of a new regulatory mechanism applicable to the most powerful AI models—those classified as "frontier." The regime requires developers to undergo state assessment and obtain official approval before broadly releasing a product. By design, it is analogous to the licensing process in aviation or pharmaceuticals—only for generative AI.

GPT-5.6 underwent this review for several weeks. During all this time, the model was accessible only to a limited circle: approximately 20 organizations received closed preview access. The names of partners are not disclosed.

• GPT-5.6 is the most advanced of OpenAI's deployed models as of July 2026
• Period of limited access: several weeks of state review
• Number of organizations in closed preview: approximately 20
• Legal mechanism: new U.S. oversight regime for frontier AI systems
• Outcome: approval received, broad release unblocked

Precedent for OpenAI's competitors

GPT-5.6 became the first model to publicly pass through the new American oversight regime—from limited preview to broad release. This sets a precedent guiding the entire industry: now it is clear the procedure takes several weeks and includes state review involving a closed group of testers.

For Anthropic, Google DeepMind, xAI, and other companies developing frontier systems, this means: the next major release will follow the same path. Time from "model is ready" to "model is available to all" now includes a mandatory regulatory pause—and for GPT-5.6, its duration was several weeks.

Details of the process itself—assessment criteria, list of mandatory tests, composition of experts—remain opaque. GPT-5.6's passage through it creates the first public reference points.

What will users get

After receiving approval, OpenAI is entitled to open GPT-5.6 to its entire audience—through ChatGPT and API. Functionality unavailable to regular users all this time becomes publicly accessible.

Specific conditions: pricing plans, regional restrictions, exact timelines for phased opening—are not specified in the source. The regulatory barrier is removed; the rest is at OpenAI's discretion.

What this means

Approval of GPT-5.6 is the first public confirmation that the new American AI oversight regime works: the model passed it and reached the market. For OpenAI, this is a victory. For the industry—a signal: the era of instant releases without state participation has ended. Each next flagship will now pass through state scrutiny before reaching users.