OpenAI released its GPT-5.6 family, comprising the Sol, Terra and Luna models, to the general public on Thursday, roughly twelve days after the U.S. Commerce Department confined access to about twenty vetted partners pending additional review. The staggered rollout, first previewed in late June, is the most concrete example yet of a de facto federal preclearance regime for frontier AI, operating without statutory basis or published procedure.

The framework traces to an executive order signed by President Trump that established a voluntary review window of up to 30 days for “covered frontier models” before broad release. Sam Altman, according to Axios, discussed the models with the White House in early June. Sol, tuned for work in biology, chemistry and cybersecurity, is the reason regulators paid attention: OpenAI says the model is “better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks,” and that its capabilities don’t reach the “critical” threshold set out in the company’s own preparedness framework.

That last claim is doing a lot of work. It’s simultaneously a reassurance to Washington and an argument that the review process shouldn’t have been necessary.

OpenAI has made the argument explicitly. In a June statement, the company said, “We don’t believe this kind of government access process should become the long-term default,” adding that “It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.” Axios has reported the company views the current moment as an “in-between period.” On Thursday, OpenAI said that “as part of our ongoing engagement with the U.S. government, we previewed our plans and the models’ capabilities ahead of today’s launch,” and described the initial preview partners as those “whose participation has been shared with the government.”

The pattern isn’t confined to one lab. Anthropic’s Fable and Mythos models were suspended for close to three weeks under an export-control directive before access was restored in late June. Two frontier labs, two administrative pauses, one improvised regime.

Sol is priced at $5 per million input tokens and $30 per million output tokens, with Terra and Luna cheaper. OpenAI says it’s working with the government to develop a “repeatable process for future model releases,” which is the diplomatic phrasing for what has effectively already emerged. The company would prefer the improvisation harden into rules it can plan around; Washington, for now, seems content with the leverage that comes from not writing them down.

Sources