Trump asks OpenAI to hold back GPT-5.6: model will go only to select partners

The White House has directly influenced the timing of a commercial AI model release for the first time: at the request of Donald Trump's administration, OpenAI is moving GPT-5.6 to a limited-distribution mode instead of a broad public launch.

What happened with GPT-5.6

According to The Information, OpenAI received a direct request from the administration of the US president to delay the public release of its new model. Officially, the request is motivated by "security reasons"; no specifics have been disclosed so far.

In response, the company switched to a phased rollout: GPT-5.6 will first go to a limited circle of "trusted partners" rather than a broad audience. Who exactly will be on this closed list is unknown. It is also unclear whether a broad release is planned at all, or whether the model will remain a closed-access tool. OpenAI has not officially commented on the request itself or on its further release plans.

The first public precedent

Until now, American AI companies set their own model release schedules. OpenAI, Anthropic, Google and Meta* relied on their own internal safety evaluations, competitive dynamics in the market and investor demands. No direct intervention by the US executive branch in this process had previously been publicly recorded.

The key detail: this is an informal "request," not a regulatory order or a legislative requirement. Even so, OpenAI complied. That suggests the company treats the White House's position as binding, even without legal compulsion.

It is also telling that all of this is happening amid a long-running debate over who should control the release of the most powerful AI systems, and how. Part of the expert community has long insisted on mandatory government vetting of frontier models before public release. The Trump administration's actions, though informal, move in exactly that direction.

GPT-5.6 in the context of the update race

OpenAI has sharply accelerated updates to its model lineup over the past several months. In a short span, the following were released:

• o3 and o4-mini: specialized models for deep-reasoning tasks
• GPT-4.5: with expanded conversational capabilities
• GPT-4.1: with a larger context window
• GPT-5: a new general-purpose flagship

GPT-5.6 was supposed to continue this chain. Its specific characteristics and differences from GPT-5 have not yet been disclosed, though industry observers suggest it may bring improvements for agentic tasks or new reasoning modes. For developers building products on the OpenAI platform, the delay adds uncertainty to planning.

What this means

The White House has de facto gained leverage over the production schedule of leading AI labs, and has used it. For the whole industry this is a landmark precedent: government oversight of AI can be exercised not only through legislation and regulators, but also through direct informal arrangements with companies.

OpenAI, for its part, has received another reminder that its position as the world's most influential AI company is inseparable from the political context, and that this link will only strengthen.

*Meta has been designated an extremist organization and is banned in Russia.