Trump Administration Lifts Testing Restrictions on OpenAI's GPT-5.6, Broader Access Expected

Since the Trump administration placed export controls on Anthropic and OpenAI models in early July, the policy environment around advanced AI has moved fast. The latest development: the administration has lifted the testing restrictions it had imposed on OpenAI's GPT-5.6, according to Axios.

What the Restriction Was

Executive Order 14409 mandated a 30-day government pre-testing phase for GPT-5.6 before any broader rollout. During that window, access was limited to government-vetted partners only. The model includes what OpenAI has internally called the flagship Sol model.

That holding period is now over.

What Changes

With the restrictions lifted, the path is open for OpenAI to move toward a wider public release, potentially on platforms like ChatGPT and Microsoft Copilot. Axios reported the policy shift, though neither OpenAI nor the White House had issued a formal public statement as of the time of that report.

No specific release date for a public GPT-5.6 rollout has been announced. Observers are watching OpenAI's official blog and ChatGPT release notes for next steps. Any public statement from CEO Sam Altman would carry weight on timing.

Why the Pre-Testing Phase Existed

Frontier AI models carry real national security implications, and a 30-day vetting window is a modest ask before something with GPT-5.6's capabilities goes fully public. OpenAI disclosed just last week that Chinese-linked operators had already used ChatGPT in a covert influence campaign targeting U.S. data center policy, so the concern about adversarial misuse of advanced models is not hypothetical.

Critics of the review requirement argued the opposite: that government gatekeeping of commercial AI releases creates competitive disadvantages for U.S. firms, especially against Chinese AI developers who face no equivalent domestic friction. Both concerns are legitimate. The administration's decision to lift the restriction suggests it concluded the vetting was complete, not that the security rationale was wrong.

The Export Control Backdrop

This development sits in some tension with the export controls the administration placed on OpenAI and Anthropic models in early July. Those controls restrict how the most capable models can be accessed outside the United States. Lifting the domestic pre-testing restriction expands access at home. The export controls simultaneously tighten the perimeter abroad.

This two-track approach, open domestically and controlled internationally, is consistent with the administration's stated China-competition framework. Whether it's the right calibration is a genuine open question. Overly tight domestic controls slow U.S. commercial deployment; overly loose international controls risk transferring capability to adversaries.

What OpenAI Has Not Said

OpenAI has not publicly confirmed a specific GPT-5.6 public launch date. The company has not issued updated safety evaluation results tied to the end of the government review period. Sam Altman has made no on-record statement about post-restriction release plans as of July 8, 2026.

The gap between a government restriction being lifted and a company actually shipping to the public can be significant. OpenAI still controls its own rollout pace, and additional internal safety reviews or commercial readiness checks could delay public availability independent of the administration's policy decision.

The concrete next marker to watch: whether OpenAI publishes a GPT-5.6 launch announcement or updated system card in the days following the lifted restriction, which would indicate the company is moving immediately rather than waiting.