On June 3, 2026, OpenAI CEO Sam Altman addressed reporters following a meeting with Hakeem Jeffries, the U.S. House Minority Leader, on Capitol Hill in Washington, D.C. During this occasion, OpenAI disclosed the launch of three novel artificial intelligence models and confirmed its adherence to the U.S. government's request to initially restrict access to a select group of trusted partners.
In a blog post released Friday, the organization emphasized its belief in the importance of widespread access to technology and articulated plans to make the models—designated as GPT-5.6 Sol, Terra, and Luna—available to the public in the upcoming weeks. OpenAI shared its insights regarding the models' functionalities and its rollout strategy with government authorities prior to the recent announcement.
OpenAI expressed its concern that the current government-access process should not become a standard operating procedure, which they argue would hinder developers, users, businesses, cybersecurity professionals, and global collaborators from utilizing some of the most advanced tools available.
However, the company has not revealed the identities of the partners who will initially access these new models.
This rollout announcement follows closely on the heels of competitor Anthropic, which recently had to suspend access to two of its advanced models in compliance with an export control directive issued by the Trump administration. Anthropic is currently engaged in discussions with Washington officials but has not specified when its models may be reinstated.
Under the Trump administration, there has been a noticeable increase in the government’s involvement in AI regulation, particularly after President Trump signed an executive order earlier this month. This order, while lacking specific instructions, urged AI developers to voluntarily permit governmental reviews of model capabilities before their full release.
OpenAI is actively collaborating with the Trump administration to help develop a framework for these evaluations and to establish a systematic approach for future model introductions.
The company stated, "We are adopting this temporary measure as we believe it is the most effective route to ensuring broader access in the near future."
The newly introduced models, GPT-5.6 Sol, Terra, and Luna, are categorized based on their capabilities, with Sol representing the most advanced model to date. OpenAI noted that this model demonstrates significant enhancements in coding and biological applications, claiming it to be the most proficient tool for cybersecurity purposes. The company highlighted that it excels in assisting users in remedying vulnerabilities, rather than executing complete cyberattacks, and remains below the threshold defined as “critical” cybersecurity risk, which involves introducing unprecedented dangers.
In summary, OpenAI is limiting the rollout of GPT-5.6 following a communication from the Trump administration.
