OpenAI's advanced models have gone live for government use
Samuel Boivin/NurPhoto via Getty Images
OpenAI’s latest GPT-5.6 models were made available for government work via the FedRAMP-authorized ChatGPT Enterprise offering, weeks after one of the advanced models was involved in executing an autonomous digital attack.
Federal agencies have gotten access to OpenAI’s latest advanced models, with the ChatGPT-5.6 series — Sol, Terra and Luna — available through the ChatGPT Enterprise suite, authorized under the Federal Risk and Authorization Management Program.
An OpenAI spokesperson and a source familiar with the models’ availability said that the GPT-5.6 series were made available under OpenAI’s existing FedRAMP–accredited product. The OpenAI spokesperson told Nextgov/FCW that the GPT-5.6 model family was made available to federal customers one day after they were released to the general public on July 9th. Agencies that are currently enrolled as ChatGPT Enterprise customers through federal programs like OneGov are able to access GPT-5.6 models.
Several of OpenAI’s 5.6 models are available to all staff within the Department of Health and Human Services via its purchase of OpenAI products via OneGov, the source familiar with the availability told Nextgov/FCW. The same source confirmed that GPT Sol is available for HHS employees “at various reasoning levels.” The models were made available to the agency in late July.
OneGov was launched in April 2025 and is managed by the General Services Administration to bring advanced technology services to the public sector at temporary discounted rates. Technology providers who wish to participate in OneGov are required to achieve FedRAMP certification.
As with other products included in the ChatGPT Enterprise suite, the GPT 5.6 series are able to handle government workflows that fall within FedRAMP’s Moderate clearance level, which covers controlled, unclassified information.
OpenAI hinted at the inclusion of the GPT-5.6 model family in their broader ChatGPT Enterprise offering — which received FedRAMP Moderate authorization in April — in an updated release that clarified that the latest model family is generally available to use in FedRAMP API, including mini and nano variants.
“FedRAMP Certifications apply to the cloud service offering listed on the FedRAMP Marketplace, including OpenAI’s ChatGPT Enterprise and API Platform,” the source familiar with the matter told Nextgov/FCW.
A government-specific version of ChatGPT first received FedRAMP certification in January 2025.
The availability of the ChatGPT-5.6 models for government customers follows an automated attack that a combination of OpenAI models — including GPT-5.6 Sol — conducted on Hugging Face networks following a breach from a testing environment. The incident has sparked renewed fears of advanced AI tools’ capabilities, with GSA FedRAMP Director Pete Waterman saying that AI developers need to be steadfast in allocating robust cybersecurity resources when developing and testing new models.
“If that is the way that you are approaching information security today … the way your company is approaching information security today, I don’t want you in the federal marketplace, and you shouldn’t be selling your software to anyone,” Waterman said on July 23 at Carahsoft’s FedRAMP Summit in Washington, D.C.
The GPT-5.6 models were unveiled in late June to a select group of partners for initial deployment testing, which included “engagement with the U.S. government.” The GPT-5.6 series are designed for different tasks and workloads, with the most advanced model, GPT-5.6 Sol, intended for software coding and development, biological and cybersecurity work, and more.