OpenAI launched a US-only preview of its GPT-5.6 model series on Friday to a limited group of partners whose identities have been shared with US authorities, according to The Hindu/AFP. The restricted release followed a government request tied to a voluntary federal review protocol established by a Trump executive order earlier this month.
The company simultaneously published a blog post pushing back on the arrangement: “We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.”
Three Models, One Restricted Launch
GPT-5.6 comprises three models. Sol is the new flagship. Terra is a mid-range model for everyday work, priced at half the cost of its predecessor GPT-5.5. Luna is a fast, low-cost option. All three are currently available only to US-based partners, though overseas employees at those partner organizations also have access, according to The Hindu/AFP.
OpenAI described the restricted launch as a short-term step toward broader availability “in the coming weeks.”
Government Review Becomes Standard Practice
The release comes two weeks after the White House ordered Anthropic to ban all foreign nationals from accessing its Fable 5 and Mythos 5 models, citing national security concerns. Anthropic shut down all access to those models entirely, stating it could not reliably comply with the restriction. NCT has covered the Anthropic export control saga extensively, including the partial restoration of Mythos 5 access to approximately 100 trusted organizations on June 27.
The core concern driving government intervention is the same across both companies: frontier models’ reportedly unprecedented ability to identify software vulnerabilities. Trump’s executive order established a voluntary federal review of national security risks in advanced AI models before release. The White House has communicated little about enforcement specifics or which models fall under the review, according to AFP.
The Commercial Pressure
The intervention is striking for a White House that has otherwise pushed to loosen AI oversight, including moves to block states from writing their own rules. OpenAI said it was “uncomfortable” with the process, but chose compliance as the fastest path to broader release.
Both OpenAI and Anthropic have filed confidential IPO documents with US regulators, targeting public listings at valuations approaching $1 trillion, according to AFP. Terra’s pricing at half of GPT-5.5 signals aggressive cost competition even as government restrictions limit initial distribution. Every week of restricted access is a week competitors can capture enterprise contracts.
What Agent Builders Face
For teams building on OpenAI’s latest models, the takeaway is operational: announced release timelines are no longer reliable. The gap between “model exists” and “model is available to you” now includes a government review phase with no published timeline or criteria. Builders who locked deployment plans to GPT-5.6 availability are discovering that model selection has become an infrastructure dependency with political risk attached.