OpenAI says it will restrict new AI model access to “trusted partners,” citing U.S. government request

OpenAI said on June 26, 2026, that it is limiting access to newly developed AI models to “trusted partners,” after what the company described as a request from the U.S. government. The company also stated that it previewed the models’ capabilities with the government ahead of the launch, according to the report.

In a statement referenced by CNBC, OpenAI did not describe the specific categories of partners in the available description, but characterized the approach as a way to manage distribution of the new models while authorities reviewed or assessed the capabilities prior to broader availability.

The restriction is tied to how OpenAI intends to roll out the new systems following the pre-launch preview. Under the company’s stated approach, access would be channeled through partners that have been vetted, rather than being immediately opened to the general public.

The reported policy raises questions for the compliance and public-interest side of AI deployment, because “trusted partner” arrangements can affect who can test the systems, under what terms, and how quickly independent developers and researchers outside the partner network can evaluate performance, safeguards, and potential misuse.

For policymakers and regulators, the reported sequence highlights an emerging operational model in which U.S. government review or coordination can influence the timing and distribution mechanics of advanced AI releases. It also underscores the role of pre-launch engagement in how companies manage launch timelines and access pathways for capabilities they view as sensitive.