OpenAI Delays GPT-5.6 Launch at White House Request

On June 26, 2026, OpenAI announced it had delayed the public release of its latest frontier AI models, the GPT-5.6 series, at the request of the Trump administration. Although these models have the potential to aid cybersecurity defenders, the company decided to offer only a limited preview to "a small group of pre-approved, trusted government partners" before wider access. The series includes three models: Sol, Terra, and Luna. OpenAI stated it plans to gradually roll out full public availability "over the coming weeks."

In a press release, OpenAI emphasized: "We believe in the value of broad access and are in ongoing conversations with the U.S. government. Prior to today's announcement, we provided officials with a preview of the models' capabilities and release plans." This delay comes amid heightened U.S. government scrutiny of AI safety. Two weeks earlier, the White House issued export control directives to Anthropic, forcing the company to take its most advanced AI model offline for all users, with no resolution yet. Some Anthropic employees remain prohibited from using the company's strongest model.

Industry analysts view OpenAI's compliance with the White House as a sign that U.S. policy is increasingly tightening the balance between AI capability proliferation and national security. As models like GPT-5.6 with complex offensive and defensive cyber capabilities emerge, federal oversight of frontier AI is expected to intensify.