OpenAI will reportedly stagger the GPT-5.6 release after the US government raised security concerns, limiting who can reach the model first.
Federal officials would gain a say over which customers receive early preview access, according to a new report.
What the GPT-5.6 Release Report Says
The Information reported that the Trump administration asked OpenAI to phase the launch rather than open it widely at once. The outlet said federal reviewers would approve preview access one customer at a time during the early window.
Staggered launches already sit in OpenAI’s playbook. The company withheld the full GPT-2 model for roughly nine months in 2019 over misuse fears. Its GPT-5.5 model launch on April 23 reached paid tiers before free users.
More directly, OpenAI shipped a cyber-focused version of GPT-5.5 only to vetted defenders under a trusted-access program. The GPT-5.6 plan would extend that template to Washington itself.
A Federal Review Framework Takes Shape
The reported request maps onto Executive Order 14409, which President Donald Trump signed on June 2. It asks developers to give the government up to 30 days of access to their most capable models before release.
Federal officials would also help choose which trusted partners get early access.
A classified benchmark led by the National Security Agency would decide which systems count as covered frontier models. The threshold turns on a model’s advanced cyber capabilities.
A separate Treasury-run clearinghouse would hunt and patch software flaws, extending the administration’s cyber defense doctrine.
The framework is voluntary and bars any licensing regime, part of a wider federal AI policy push. Officials cast it as a way to test frontier models for cyber risks. Some former advisers have criticized that case as overblown.
OpenAI has not officially confirmed GPT-5.6 or a firm launch date, and earlier timing has slipped toward July. How tightly Washington shapes early access could set a template for the next frontier releases from OpenAI and Anthropic.
Bitrue