OpenAI Limits New AI Models After U.S. Government Request

President Donald Trump signed an AI executive order earlier this month. The order asked AI developers to voluntarily let the government assess model capabilities before a full release. It was thin on specific details.

OpenAI said it is working with the administration to help build a framework for such assessments and to create a repeatable process for future model releases.

"We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks," OpenAI said.

GPT-5.6 Sol is the company's strongest offering yet, according to OpenAI. It shows improvements in coding and biology and is the company's most capable model for cybersecurity. OpenAI said the model is better at helping users fix vulnerabilities than at carrying out end-to-end attacks. It still does not cross OpenAI's critical cybersecurity risk threshold, which the company defines as bringing "unprecedented new pathways to severe harm."

The delay in general availability creates a near-term revenue risk for OpenAI, which relies on enterprise and developer subscriptions. Competitors like Anthropic face the same bottleneck. The administration's approach – voluntary pre-release review with no fixed timeline – leaves AI companies guessing when they can fully monetize new models. A clearer framework would reduce that uncertainty. Tighter controls or export restrictions would make it worse.

The next concrete marker is the framework OpenAI and the administration are developing. If it speeds up approvals, the current bottleneck fades. If it becomes a permanent gate, the competitive dynamics shift toward companies that can navigate the process fastest.