White House Pressures OpenAI to Delay Next-Gen AI Model Launch

According to reports from VitalLaw and CNN, OpenAI announced on June 26 that, at the request of the Trump administration, it will temporarily withhold the public release of its latest frontier AI model series, GPT-5.6, even though these models are considered beneficial for cybersecurity defense. OpenAI stated that it will first provide a limited preview to "a small group of trusted partners whose names have been shared with the government," involving three models: Sol, Terra, and Luna. The company's statement said: "We believe in broad access and plan to fully open the GPT-5.6 series in the coming weeks… As part of our ongoing collaboration with the U.S. government, we briefed them on our plans and model capabilities ahead of today's announcement."

The restriction stems from White House concerns over model safety. Earlier, the government imposed export controls on Anthropic, requiring it to withdraw the Mythos and Fable models. A source familiar with the matter indicated that OpenAI's latest models are considered comparable to Anthropic's Mythos in capability. In an internal memo, OpenAI CEO Sam Altman said this is not the company's long-term preferred approach, but given the lack of a federal AI regulatory framework, they agreed to the release restriction as a transitional step toward broader access. The move underscores the policy tug-of-war between capability and safety for advanced AI models.

Editor's Take: The shift from "voluntary restraint" to "government intervention" shows U.S. AI regulation tilting from industry self-governance toward administrative oversight. In the short term, this buys time for safety evaluations, but in the long run, the absence of clear legislation could lead to inconsistent rules across different companies. Industry leaders and lawmakers need to establish a sustainable, transparent framework as soon as possible.