OpenAI launched its most powerful AI yet on Friday, June 26 — but most people can't use it. The company said it is limiting access to GPT-5.6 at the request of the US government, according to Reuters, which reported that OpenAI is deferring a full public launch and restricting initial access to a small group of vetted partners.
The new lineup includes three models — Sol, the flagship, plus Terra and Luna, according to Axios and The Verge. Axios reports the limited preview is going to roughly 20 companies, with participants disclosed to the US government. OpenAI told CNBC it previewed the models' capabilities with the government before launch, and said it is complying as part of its agreement with the Department of Defense, which lets the department use its AI models.
OpenAI framed the restriction as a one-time accommodation, not a precedent. "We don't believe this kind of government access process should become the long-term default," the company said, per TechCrunch, adding that it "keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them." According to the Wall Street Journal, OpenAI hopes to make GPT-5.6 generally available in the coming weeks.
The move follows mounting government concern over advanced AI. Wired reports the White House asked OpenAI to delay the rollout roughly two weeks after rival Anthropic had to take its most advanced models offline. The American Bazaar and others tie the request to a cybersecurity review. On capability, The Decoder reports OpenAI's GPT-5.6 Sol beats Anthropic's Claude Mythos 5 in coding benchmarks; OpenAI's own preview cites gains in coding, science, and cybersecurity. The news rippled into markets, with Yahoo Finance UK reporting the FTSE 100 fell as the delay knocked tech stocks.
Why it matters: For the first time, the US government is effectively deciding who can use a leading commercial AI model first — a striking shift in who controls access to frontier technology.