OpenAI has restricted the initial release of its newest artificial intelligence model, GPT-5.6 Sol, after a request from President Donald Trump’s administration, marking a significant moment in the growing relationship between advanced AI companies and government oversight.

The model, described by OpenAI as its most capable system to date, is currently available only to a small group of trusted partners and organizations. Individual users and the general public do not have access during the preview period. OpenAI says the phased launch is temporary and that broader availability is expected in the coming weeks.

GPT-5.6 Sol is part of a new model family that also includes GPT-5.6 Terra and GPT-5.6 Luna. Sol is positioned as the flagship model, with stronger performance in coding, scientific work, professional tasks, and cybersecurity-related analysis. Terra is designed as a lower-cost balanced model, while Luna is intended for faster and more economical use cases.

The restricted rollout comes amid heightened concern in Washington over the national-security implications of frontier AI systems. Advanced models are increasingly capable of assisting with software development, vulnerability research, automated reasoning, and complex technical workflows. Those same strengths can raise concerns about misuse, especially in cybersecurity.

OpenAI said it previewed the model’s capabilities to the U.S. government before launch and agreed to begin with a limited group of approved partners. The company also emphasized that the model has been released with its strongest safety measures so far, including additional safeguards against prohibited cyber activity and misuse.

At the same time, OpenAI signaled discomfort with the idea that government access approval should become a permanent standard for AI launches. The company argued that broad access to powerful AI tools is important for developers, businesses, researchers, and cyber defenders, especially those using AI to find and fix vulnerabilities.

The episode highlights a central tension in the AI industry: how to balance innovation and public access with security, accountability, and government oversight. For policymakers, the concern is that highly capable AI models could accelerate cyber threats if released too broadly or too quickly. For AI companies, the risk is that heavy-handed restrictions could slow adoption, weaken competitiveness, and limit legitimate defensive uses.

For now, GPT-5.6 Sol remains in a controlled preview. OpenAI says it will continue testing the model, coordinating with partners, and working with the administration on a repeatable framework for future model releases.

The decision may become a precedent for how the United States handles the release of next-generation AI systems. As models become more powerful, the question is no longer only what they can do, but who should be allowed to use them first — and under what conditions.