OpenAI Restricts GPT-5.6 to Pre-Approved Partners

OpenAI built three powerful new AI models. But the company is limiting their release to a "small group of trusted partners" following a U.S. government request. The administration has moved to keep all three models from general release. That creates a tension between the company's desire to reach customers and the government's push to control advanced AI.

This move reflects the escalating regulatory pressure on frontier AI developers. The Trump administration has increasingly demanded that companies restrict the release of advanced systems, citing national security concerns. The situation mirrors Anthropic's recent experience with its Fable 5 model, which was pulled entirely after the government ordered a cutoff of access for foreign nationals. Such actions highlight the growing conflict between innovation and control, especially as the U.S. seeks to maintain a competitive edge over China in the AI race.

The GPT-5.6 family consists of three variants: the flagship Sol, the versatile Terra for everyday tasks, and the budget-friendly Luna. The flagship Sol is priced at $5 per million input tokens and $30 per million output tokens; Terra costs half those rates, while Luna runs $1 per million input tokens and $6 per million output tokens. OpenAI says Sol achieves the same performance using only one‑third of the output tokens that Anthropic's Claude Mythos 5 requires.

Get your free investing masterclass bonus when you join Market Briefs, our free daily newsletter

The Government's Role

The Trump administration has escalated demands for AI firms to limit their most cutting‑edge systems. After Anthropic launched its strongest public model, Fable 5, the government directed the company to cut off access for any foreign national, which led Anthropic to pull the model entirely.

Dean Ball, who previously served as a White House AI adviser and will soon join OpenAI, said, "President Trump's recent executive order, which asks certain AI companies to voluntarily submit their most advanced models for government review up to 30 days before release, has created a de facto involuntary licensing regime for frontier AI, leading to heavy‑handed restrictions." Ball added, "When the government doesn't have clearly defined safety standards, it could lead to endless launch delays that might not only give a hand to China in the AI race but also jeopardize the billions of dollars going to AI infrastructure buildouts."

What's Next

OpenAI described the preview as a "short‑term step" that paves the way for wider availability in the coming weeks. In a Friday blog post, the company said it will collaborate with the administration on crafting a new cybersecurity executive order framework and a "repeatable process for future model releases." The company stated: "We don't believe this kind of government access process should become the long-term default. "It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them"."

Worth Noting

OpenAI also says Sol comes with its most advanced security stack to date, designed to resist adversarial attacks and prioritize defensive cybersecurity. The safety measures are integrated into the model's fundamental behavior, not added as an external filter - an approach intended to avoid the false-positive issues that plagued Anthropic's Fable 5, where the model's classifiers would route safe prompts to an older model, causing many false positives and user backlash.

Subscribe to Market Briefs, our free daily newsletter, and claim your bonus investing masterclass