OpenAI stalls GPT-5.6 public release; US govt to approve access customer-by-customer

OpenAI will delay the general release of GPT-5.6 after the Trump administration requested the company limit initial access to a small group of government-approved enterprise partners. Sam Altman told employees on June 25 that the Office of the National Cyber Director and Office of Science and Technology Policy will vet customers individually during an extended preview period. Commerce Secretary Howard Lutnick warned Altman against any public release without cross-agency approvals. Altman said in an internal memo that he hopes a wider rollout will happen "a couple of weeks" after the limited preview concludes, but the timeline now depends on government case-by-case clearances.

The arrangement mirrors restrictions placed on Anthropic's Mythos and Fable 5 models earlier in June, which were taken offline after export-control directives. Altman made clear the government-gated access approach is "not our preferred long-term model," but OpenAI is cooperating voluntarily. No mandatory pre-release approval framework exists yet—President Trump's earlier executive order calls for a voluntary 30-day government review window, with this GPT-5.6 arrangement being the first major industry test of that process.

For architects and product leads: this sets a new de facto precedent. Frontier model releases now face indefinite gating by national-security reviews with no published criteria. Early access will concentrate on government-vetted enterprise partners, delaying third-party testing and integrations. The unpredictability around customer approval timelines introduces friction into deployment roadmaps, especially for teams that do not have direct government relationships or already-cleared status.