OpenAI limits GPT-5.6 rollout after government request, says restrictions shouldn’t be the norm

OpenAI has announced a limited rollout of its latest AI models, including the flagship GPT-5.6, in response to a request from the U.S. government. This rollout is confined to a select group of trusted partners, as the administration enforces new measures aimed at controlling access to advanced AI technologies. The newly unveiled GPT-5.6 lineup features three distinct models: Sol, the most powerful of the trio; Terra, designed for everyday use; and Luna, which offers a cost-effective solution. Despite Sol's capabilities, the Trump administration has imposed restrictions on the release of all three models, with access granted only to partners whose involvement has been disclosed to the government. This development comes as the U.S. government intensifies pressure on AI companies to manage the deployment of their most sophisticated systems. A recent incident involving Anthropic, which was forced to withdraw its Fable 5 model from public access following government intervention, has sparked discussions regarding the extent of governmental authority over AI model launches. Dean Ball, a former advisor on AI at the White House and an incoming OpenAI employee, highlighted a concerning trend: the administration’s executive order requiring AI firms to submit advanced models for government review up to 30 days prior to release effectively creates an involuntary licensing system. This has led to significant delays and could hinder the U.S. in the global AI competition, as well as threaten investments in AI infrastructure. While OpenAI complied with the government's request this time, the company expressed dissatisfaction with the arrangement. In a recent blog post, OpenAI stated, "We do not believe this kind of government access process should become the long-term default, as it restricts access to essential tools for users, developers, and security professionals." OpenAI characterized the current preview of GPT-5.6 as a temporary measure, indicating plans to expand access in the coming weeks. The company is also collaborating with the government to establish a new framework for future AI model releases and cybersecurity protocols. The GPT-5.6 Sol model boasts advanced capabilities in coding, biology, and cybersecurity. It incorporates a “max” reasoning effort mode and an “ultra” mode, which enables the model to tackle highly complex tasks more efficiently. In terms of performance, OpenAI claims that GPT-5.6 outperforms Anthropic's Claude Mythos 5 in coding workflows, despite facing similar governmental restrictions. The Sol model is noted for its robust security features, designed to enhance defensive cybersecurity measures while minimizing vulnerabilities to adversarial attacks. OpenAI is aiming to avoid pitfalls seen with models like Fable 5, which encountered issues with excessive caution in handling sensitive topics. As the GPT-5.6 models are initially released to a limited audience, OpenAI intends to broaden access to users of ChatGPT, Codex, and the API in the near future. The new models come in various sizes with tiered pricing, catering to different user needs and budgets.