SAN FRANCISCO—OpenAI has delayed the full public release of its GPT-5.6 artificial intelligence models after a request from the U.S. government, limiting initial access to a small group of vetted partners while federal officials evaluate potential national security risks associated with frontier AI systems, the company said.
The ChatGPT developer said the restricted rollout was a temporary measure taken as part of ongoing discussions with Washington over a framework for future releases of advanced AI models. The company said participating organizations in the preview had been shared with the U.S. government before broader availability.
“We don’t believe this kind of government access process should become the long-term default,” OpenAI said in a public statement announcing the launch. The company added that it was pursuing the limited preview because it believed it offered the strongest path toward wider availability in the coming weeks while discussions on cybersecurity review procedures continue.
The move follows a request from the administration of President Donald Trump for OpenAI to restrict access to GPT-5.6 during an initial review period, according to OpenAI and people familiar with the matter cited by multiple news organizations. The request comes as U.S. officials seek earlier access to frontier AI systems to assess risks including cyberattacks and potential military misuse before broader deployment.
OpenAI said the GPT-5.6 family includes three models—Sol, Terra and Luna—designed for different performance and cost requirements. The company said the models incorporate expanded safety protections and improvements in coding, reasoning and cybersecurity capabilities.
The restricted launch marks one of the clearest examples to date of direct U.S. government involvement in the rollout of a commercial frontier AI model. Industry observers have said the arrangement raises broader questions about how advanced AI systems should be reviewed before public release and what role governments should play in determining access.
The decision also follows recent government scrutiny of other leading AI developers, including Anthropic, whose latest frontier models have faced access restrictions amid cybersecurity concerns, according to company statements and media reports.
OpenAI said the limited preview is intended to be temporary and that it plans to expand access to GPT-5.6 through ChatGPT, its API and other products after completing the current review process. The company said it is continuing to work with U.S. officials on establishing what it described as a repeatable framework for future frontier AI model releases.
