OpenAI Delays GPT-5.6 Release, Will Restrict Initial Access to Government-Approved Customers

OpenAI will delay the public release of its GPT-5.6 model and restrict initial access to customers approved by the US federal government, CEO Sam Altman told employees in a company memo obtained by The Information. The decision follows requests from the White House amid national security concerns.

According to The Information, Altman said the government will be "approving access customer by customer during this preview period." A wider release is expected "a couple of weeks later," Altman said, but he stressed that the company does not view government-controlled rollouts as a long-term solution.

Government involvement and timeline

The request came from the Office of the National Cyber Director and the Office of Science and Technology Policy, with Commerce Secretary Howard Lutnick also involved. President Donald Trump signed an executive order earlier this month asking AI companies to voluntarily submit powerful models for federal review before public release.

The move follows a separate incident where OpenAI rival Anthropic disabled access to two models after a federal directive that blocked foreign nationals from using them, without specifying underlying security concerns. The GPT-5.6 preview period is expected to last two weeks.

• Sam Altman told employees in a staff memo that the staggered release is not OpenAI's preferred model, but they will work with the government for a more sustainable approach.
• The government review framework is still being standardized, creating confusion about how voluntary the process actually is.
• Neither the White House nor the Office of the National Cyber Director responded to The Information's requests for comment.
• The Information first reported the internal memo and Altman's Q&A with staff.

Implications for AI safety regulation

The delay underscores a shift from voluntary industry self-regulation toward direct government gatekeeping of advanced AI capabilities. OpenAI has previously advocated for regulatory guardrails but has resisted customer-by-customer approval, which Altman called unsustainable.

Industry observers say the episode signals that US agencies are moving from oversight requests to operational control over model releases. The administration's executive order on AI review, though nominally voluntary, is now backed by de facto enforcement via access restrictions. As the federal review framework takes shape, companies face uncertainty over timelines and criteria for approval.

What comes next for OpenAI is a phased rollout beginning with government-cleared partners, followed by a broader public release once the two-week preview period ends. Altman's memo made clear the company will seek a more systematic review process for future models.