OpenAI restricts GPT-5.6 rollout following government request, warns against normalization
OpenAI has limited the release of its new GPT-5.6 models to a small group of trusted partners following a request from the U.S. government.
According to TechCrunch, OpenAI has limited the release of its newest AI models to a small group of trusted partners following a request from the U.S. government. The next generation GPT-5.6 lineup includes three variants: Sol, which is the flagship model; Terra, designed as a more balanced option for everyday use; and Luna, described as a faster and lower-cost choice. Although Sol serves as the company's most powerful system, the Trump administration restricted the release of all three models.
The administration's request arrives amid mounting pressure on artificial intelligence companies to restrict their most advanced systems. TechCrunch notes that after Anthropic released its public model Fable 5, the administration ordered the company to remove access for foreign nationals, leading Anthropic to take the model down entirely. Dean Ball, a former White House AI adviser and soon-to-be OpenAI employee, told TechCrunch that a recent executive order from President Trump has established a de facto involuntary licensing regime.
Ball explained to TechCrunch that this regime leads to heavy-handed restrictions, particularly when the government lacks clearly defined safety standards. Such uncertainty could result in endless launch delays, potentially aiding China in the artificial intelligence race and jeopardizing billions of dollars directed toward infrastructure buildouts. Although OpenAI complied with the administration's request, the firm stated in a blog post that this arrangement should not become the industry standard.
OpenAI warned that these restrictions prevent the best tools from reaching users, developers, enterprises, cyber defenders, and global partners. The company described the initial preview as a short-term step intended to place GPT-5.6 on a path toward broader availability in the coming weeks. During this period, OpenAI plans to work with the administration to develop a new executive order framework regarding cybersecurity and establish a repeatable process for future releases.
TechCrunch reports that GPT-5.6 Sol is OpenAI's strongest model to date, featuring improved agentic capabilities across coding, biology, and cybersecurity. The model introduces a max reasoning effort mode along with an ultra mode that utilizes coordinated subagents for complex tasks. OpenAI states that Sol performs slightly better at coding workflows than Anthropic's Claude Mythos 5 while using a third of the output tokens.
To address safety concerns, OpenAI integrated its most robust security stack directly into the core model's behavior rather than relying on a separate filter. The architecture is hardened against adversarial attacks and prioritized for defensive cybersecurity work. The models are initially available to select partners, with plans for broader access via ChatGPT, Codex, and the API soon, accompanied by tiered pricing structures.