White House Intervenes to Stagger OpenAI GPT-5.6 Release Over Cybersecurity Concerns

A Highly Controlled Rollout for GPT-5.6

The United States government has requested that OpenAI limit and stagger the release of its upcoming GPT-5.6 model due to advanced capabilities that have sparked national security concerns. OpenAI has agreed to the request, which will see the new model launch initially as a highly restricted preview for a select group of enterprise customers. During this initial phase, the Trump administration plans to evaluate and approve access on an individual, customer-by-customer basis. OpenAI Chief Executive Officer Sam Altman communicated this arrangement to employees in an internal memo on Thursday, June 25, 2026. While Altman noted that OpenAI agreed to the slow rollout, he explicitly stated that government-controlled customer approvals are not the company's preferred long-term release model.

Cybersecurity Risks and Parallel Restrictions

The administration's request is driven by fears over the model's potential to perform highly advanced cybersecurity work. Officials from the White House Office of the National Cyber Director and the Office of Science and Technology Policy consider GPT-5.6 to be on par with Anthropic's restricted Mythos model. The government wants multiple federal agencies to thoroughly test the model's built-in safeguards before allowing wider commercial availability. This intervention follows a similar national security action taken against Anthropic, which was forced to pull its advanced Mythos and Fable models from the market after receiving an export control order from the Commerce Department. Those models had raised serious concerns on Wall Street and in Washington over unprecedented safety risks associated with their advanced capabilities.

Regulatory Confusion in a Transitional Era

This case-by-case approval process highlights a growing challenge for frontier artificial intelligence labs operating without a clear, centralized regulatory framework. Earlier in June 2026, President Donald Trump signed an executive order establishing voluntary government testing of advanced models for up to 30 days before public release, but the official system for executing this has yet to be established. Consequently, AI companies face confusion regarding which federal entity holds regulatory authority, as the restrictions on Anthropic came from the Commerce Department, while the request to OpenAI originated directly from the White House. Although Commerce Secretary Howard Lutnick has discussed the release of GPT-5.6 with Altman, and the White House has already previewed the model, there remains no transparent, consistent system for managing these high-stakes releases.

As federal agencies scramble to police frontier models on an ad-hoc basis, the lack of a standardized regulatory framework leaves both developers and enterprise clients in a highly unpredictable limbo.