White House asks OpenAI to delay GPT-5.6 release, two weeks after Anthropic

OpenAI has built new GPT-5.6 models, but users will not get access to them for now. According to Wired, the White House asked the company to delay the public release. It is the second such case in two weeks: Anthropic found itself in a similar situation two weeks earlier.

What happened with OpenAI

OpenAI has finished developing its next generation of language models, branded GPT-5.6. The company was preparing a public launch, but the Trump administration stepped in: OpenAI received a request to hold off on the release. Neither the White House nor the company has given any public explanation, either of the reasons for the delay or of how long it may last.

For developers and enterprise customers building products on top of the OpenAI API, this creates direct uncertainty: it is unknown when exactly the new capabilities will become available. Planning product roadmaps in these conditions is considerably harder.

Anthropic faced this first

The GPT-5.6 episode follows an equally telling precedent. Two weeks earlier, Anthropic was forced to pull its most advanced AI models from public access. Again there was no official explanation; users and developers learned about the restrictions after the fact.

Two events in such a short span are a pattern, not a coincidence:

• The US government has begun to treat frontier AI models as strategic assets that require state control
• The mechanism of influence remains informal: not laws and public regulations, but direct requests to specific companies
• Neither OpenAI nor Anthropic has disclosed the nature of the requests they received or the criteria behind their decisions
• Developers learn about restrictions after the fact, with no advance warning
• The precedent is being set without broad public debate or legislative codification

Regulation without public rules

The European AI Act builds regulation on transparent, predictable requirements: companies know in advance what they must do to bring a model to market. The American approach, as seen in these two cases, works on a fundamentally different principle. Instead of clear public criteria, there are informal negotiations behind closed doors.

For businesses that depend on the APIs of leading AI companies, this creates a hard-to-predict risk: the release date of the next model is now determined not only by technical readiness but also by the White House's position.

Meanwhile, it is American companies — OpenAI, Anthropic, Google DeepMind — that hold global leadership in building the most powerful language models. Systematic release delays could weaken the competitive position of the US, especially against Chinese AI developers that enjoy active state support.

What this means

Not long ago, tech companies set the pace of model releases themselves, without looking over their shoulders at regulators. Now the White House is getting involved in that process: so far informally, with no public rules or transparent criteria. Two cases in two weeks indicate that these are not isolated incidents but the start of a new phase in the US government's relationship with the AI industry.