OpenAI said late Tuesday that GPT-5.6, its flagship Sol model alongside lower-tier Terra and Luna variants, will launch publicly on Thursday, closing weeks of voluntary technical testing coordinated with the Center for AI Standards and Innovation inside the Commerce Department. It’s the first frontier-model deployment to move through the framework established by President Trump’s June 2 executive order on AI innovation and security.

The White House was quick to correct the framing. Told that OpenAI’s rollout was being described as an administration approval, an unnamed official said “no such permission is required or granted,” adding that release-timing decisions “rest entirely with the companies.” The executive order, on its text, requires no preclearance, only optional review coordinated with CASI and, where relevant, the National Security Agency.

The optics tell a different story. A person familiar with the matter said OpenAI dispatched technical staff to Washington and kept them there to field follow-up questions, a posture that critics quoted by NPR say makes the voluntary label hard to distinguish from a mandatory one in practice. The parallel is the 2016 encryption debates, where “voluntary cooperation” between Silicon Valley and federal agencies calcified into an expected default.

The commercial subtext is louder than the regulatory one. Anthropic closed a $65 billion Series H in late May at a $965 billion valuation, jumping past OpenAI’s $852 billion March mark. A Thursday launch, cleared through Washington, is narrative management as much as product release.

Downstream, SMB teams are hedging. Model-agnostic platforms like LemonLime, Zapier and Make are among the fastest-growing options for swapping underlying models without rebuilding workflows, a quiet insurance policy against whichever lab wins the next review cycle.

Sources