News
GPT-5.6 Rolls Out to Everyone, White House Denies Approval
OpenAI made its three GPT-5.6 models, Sol, Terra and Luna, available to the general public on July 9, after a two-week restricted-access period imposed by a government security review. When Axios reported that authorities had formally approved the release, the White House denied it, exposing confusion over new AI oversight rules in the US.
On July 9, OpenAI opened full public access to its GPT-5.6 model family across ChatGPT, Codex, ChatGPT Work and the API. That closed a two-week window during which the company had to restrict its newest models to roughly twenty vetted partners at the request of the Trump administration. Hours after the announcement, the White House denied reports that it had given OpenAI the formal go-ahead, exposing just how murky the rules governing oversight of the most powerful AI models in the United States have become.
Two Weeks Under Lock
When OpenAI announced GPT-5.6 Sol on June 26, the model went only to a narrow group of partners whose identities the company disclosed directly to the administration. The reason was a June presidential executive order on AI cybersecurity, which asked companies to voluntarily submit their most powerful models to government review before releasing them to the wider public. In practice, that meant OpenAI's newest model was available for twelve days only through the API and Codex, and only to organizations approved by the government.
OpenAI received clearance for the wider launch after the Center for AI Standards and Innovation at the Department of Commerce ran additional tests and held a series of meetings with the company's engineers in Washington. On July 9, the model became available to all ChatGPT users, the API, Codex, and the corporate ChatGPT Work tier, with full global rollout expected to take another 24 hours.
Who Actually Approved It
Things got more complicated when Axios reported that OpenAI had received the formal blessing of federal authorities, which would have been the first precedent of its kind in the history of US AI regulation. White House spokespeople quickly denied that version of events to CNBC, saying that decisions about deploying the model "rest solely with the companies" and that the administration had not given OpenAI any "green light".
We don't believe this kind of government access process should become a long-term standard - OpenAI, company statement
The dispute reveals something more significant than a communication mishap. The United States still has no formal legal mechanism requiring AI companies to obtain government approval before releasing a new model. The entire process OpenAI went through rested on voluntary participation rather than a hard statutory obligation, which is exactly why it's unclear whether the two-week delay was binding at all.
The Bigger Oversight Picture
The GPT-5.6 case fits into a broader process in which the Trump administration is negotiating voluntary frameworks with OpenAI, Google and Anthropic for testing and releasing the most advanced AI models. According to earlier reports, the final version of such rules could give the government up to a 30-day window to conduct a safety review of any model deemed "frontier" before it reaches partners other than national security agencies. Anthropic has already proposed its own severity scale for rating jailbreaks, modeled on the CVSS vulnerability scoring system that has been used in the cybersecurity industry for twenty years.
For Polish companies and developers using OpenAI's API, the practical takeaway matters most: GPT-5.6 Sol, Terra and Luna are now available without restrictions, and the pricing shows clear segmentation, with Terra offering performance close to GPT-5.5 at half the cost, which could meaningfully lower bills for businesses running OpenAI models at scale. At the same time, the episode shows that the US AI market is entering a phase where political negotiations between Washington and the labs can physically delay access to new tools, even without a hard legal basis.
What Comes Next
The key question is whether the voluntary review mechanism OpenAI submitted to will become a permanent fixture of model releases in the US, or remain a one-off precedent. The proposed framework involving OpenAI, Google, Anthropic, Microsoft and Amazon is expected to be finalized in the coming weeks, and the deadline set by Trump's June executive order runs out in early August. Until then, AI companies will have to operate under a cloud of uncertainty, well illustrated by the dispute over who actually approved the GPT-5.6 launch.
Sources: OpenAI publicly release GPT-5.6 (CNBC, cnbc.com), OpenAI rolls out GPT-5.6 (Engadget, engadget.com), White House Denies Approving OpenAI's GPT-5.6 Rollout (TechBuzz AI, techbuzz.ai), GPT-5.6 public rollout starts July 9 (TechMyMoney, techmymoney.com)
