OpenAI will publicly launch its most capable model, GPT-5.6, on Thursday after delaying the release last month at the request of the U.S. government amid national security concerns over the potential misuse of powerful AI systems.
The launch comes after the U.S. government lifted restrictions on Anthropic’s latest Fable and Mythos AI models last week, less than three weeks after the company was ordered to suspend access to them over national security concerns.
OpenAI CEO Sam Altman
(Photo: Shelby Tauber/Reuters)
Washington has increased scrutiny of advanced AI model releases as officials seek to identify potential threats, amid concerns that the technology could be misused by military or intelligence agencies in China, Russia or other countries of concern.
Axios, which first reported the OpenAI launch, said a source familiar with the matter told the outlet that the U.S. Department of Commerce approved a broad release of GPT-5.6 following additional government testing under Washington’s new oversight framework for frontier AI models.
OpenAI had previously limited access to the model to a small group of vetted partners whose details were shared with U.S. authorities.
The company said in a post on X late Tuesday that it plans to launch GPT-5.6 Sol, along with two additional models, Terra and Luna.
Sol is OpenAI’s most advanced model to date, while Terra is a mid-tier, lower-cost model and Luna is the company’s most cost-efficient option.
The White House and the US Department of Commerce did not immediately respond to Reuters requests for comment outside regular business hours.
The increased scrutiny of AI models began after U.S. President Donald Trump signed an executive order establishing a voluntary framework requiring developers of “covered frontier models” to provide the government with access to their systems for up to 30 days before releasing them to trusted partners.
Anthropic has warned that making any AI model completely resistant to jailbreak attempts is “probably impossible,” noting the potential development of a universal jailbreak capable of bypassing safeguards across an entire class of harmful behaviors.