White House pressures OpenAI to slow roll GPT 5.6 release over safety

OpenAI's newest model, GPT 5.6, won't see a typical public launch. According to The Information, CEO Sam Altman told staff that the Trump administration has requested a tightly controlled release. The White House's Office of the National Cyber Director and Office of Science and Technology Policy asked that access be limited to a select group of partners, with the government approving each customer individually. Altman added that if the preview goes well, a broader release could follow in a couple of weeks. This marks a notable shift for the Trump administration, which initially advocated a hands-off approach to AI regulation but recently signed an executive order requiring voluntary safety testing for frontier models.

Anthropic has already faced controversy for its similar approach with Claude Mythos, a frontier cyber model released only through its Project Glasswing program. The company argued the model was too powerful and could cause harm if misused. Critics question whether this is genuine safety caution or a marketing tactic. However, the underlying concerns are real: LLMs can write malware, execute ransomware autonomously, and find software vulnerabilities at speeds no human can match. Since both GPT 5.6 and Claude Mythos remain closed, it's difficult to independently verify their true capabilities and risks.