OpenAI GPT-5.6 Under U.S. Government Review, Full Release Delayed On June 25, the Trump administration requested OpenAI to adjust the release plan for GPT-5.6. Instead of making it directly available to all users, access will first be granted to a limited number of government-approved partners. This adjustment comes as GPT-5.6 is believed to have capabilities approaching those of Anthropic's cutting-edge models, Mythos and Fable. The U.S. government is concerned about potential risks such as cyberattacks and cybercrime. Previously, Trump signed an AI safety executive order requiring advanced AI models to undergo government safety testing and evaluation before release. However, the regulatory framework for this is not yet fully established. Around June 24, OpenAI CEO Sam Altman discussed the GPT-5.6 release plan with U.S. Commerce Secretary Howard Lutnick. Following this, OpenAI adjusted the release timeline. The White House stated that these measures aim to improve AI safety assessments and are not intended to block the release of GPT-5.6. In an internal memo, Altman mentioned that GPT-5.6 will initially be available on a limited basis, only to government-approved clients, with plans to gradually expand access later. "We will collaborate with the government and other industry players to explore more sustainable release methods in the future.