Regulatory easing for GPT-5.6: OpenAI granted full deployment, cutting-edge AI approval system temporarily ends

OpenAI has obtained a key regulatory clearance for its flagship model GPT-5.6. This not only brings the company a step closer to full commercialization but also marks the first substantial easing of the U.S. government's “case-by-case approval” mechanism for cutting-edge AI models. On July 8, according to Axios and CNBC, OpenAI has officially received approval from U.S. regulators for the launch plan of GPT-5.6. This means GPT-5.6 will move from a restricted release limited to a handful of partners to broader commercial deployment—an important milestone for the full rollout of this series. Just two weeks ago, the GPT-5.6 series was still under strict regulation. On June 27, when OpenAI first released the GPT-5.6 Sol, Terra, and Luna models, the U.S. government required that only about 20 “trusted partners” be given limited preview access. At that time, OpenAI unusually expressed dissatisfaction in an official statement, stating that government intervention in the access process “should not become the default mode in the long term,” and pointed out that this arrangement restricted developers, businesses, and cybersecurity agencies from obtaining access to cutting-edge AI capabilities. The regulatory green light this time means that the standoff between OpenAI and the U.S. government over the deployment scope of GPT-5.6 has reached a breakthrough, making it the first case to go through the complete process from “restricted release” to full approval under the new U.S. frontier AI review framework. From "Restricted Release" to Full Approval This round of regulatory tightening began in mid-June. Just 72 hours after Anthropic launched its Mythos-level model Fable 5, it was pulled from the market under congressional pressure. The reason was that in a closed-door demonstration, it showcased potentially dangerous capabilities such as draining bank accounts, raising high concerns among U.S. policymakers about frontier AI security risks. Subsequently, U.S. Representative Andrew Garbarino called for a unified regulatory review mechanism for advanced AI models. Against this backdrop, GPT-5.6 became the first pilot case under this regulatory framework. Unlike Anthropic’s forced withdrawal, OpenAI adopted a more proactive strategy, including presenting the model’s capabilities to regulators in advance, accepting a phased release arrangement, and pledging to continue security testing and strengthen risk assessment with partners during the limited preview. Now that regulatory approval has been granted, OpenAI has not only completed the regulatory process from restricted release to full launch ahead of its peers but has also seized the initiative in commercialization. Sol Officially Debuts This Week, With Pricing Targeted at Anthropic The flagship model Sol from the GPT-5.6 series will officially launch this Thursday. OpenAI CEO Sam Altman previously revealed that Sol will offer a new Ultra multi-agent mode and Max inference strength version, with API pricing at $5 per million input tokens and $30 per million output tokens—about half the price of Anthropic Claude Fable 5. In terms of performance, on the Terminal-Bench 2.1 coding benchmark, Sol Ultra scored 91.9% and the standard Sol version 88.8%, both higher than Claude Mythos 5’s 88.0%. The lower price, coupled with equal or better performance, makes Sol poised to directly challenge Anthropic’s enterprise API market pricing system. At the same time, OpenAI plans to deploy Sol on Cerebras hardware later this month, with inference speed expected to reach 750 tokens per second, further enhancing its competitiveness in high-throughput enterprise scenarios. U.S. AI Regulation Enters a New Phase Approval for GPT-5.6 does not mean the U.S. regulatory system for frontier AI is set in stone—it is more like the first step in a new regulatory regime. OpenAI previously called on the U.S. government to establish a repeatable, predictable, and standardized approval mechanism rather than a case-by-case review for each model. Whether the smooth approval of GPT-5.6 signals the formation of a relatively fixed review process for future cutting-edge AI models in the U.S.—and whether products like GPT-6, new Anthropic models, or Google Gemini will follow the same mechanism—remains to be seen. While Anthropic is still seeking regulatory approval to relaunch Fable 5, OpenAI, by engaging more proactively with regulators, has achieved commercial clearance first. This not only gives OpenAI a timing advantage for product launches but is also likely to reshape the market’s valuation of the two companies’ commercialization execution and regulatory risk management capabilities. Risk Warning and Disclaimer The market involves risks, and investments should be made cautiously. This article does not constitute personal investment advice and does not take into account individual users’ specific investment goals, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment is at your own risk.