By submitting its models to government approval on a client-by-client basis, OpenAI cements the end of unrestricted releases and turns the State into the natural arbiter of the technological frontier.
There is a delicious irony in the fact that OpenAI, a company that sold itself to the world as an agent of unchecked disruption, has become hostage to a government calendar to launch its own products. According to The Verge, the limited release of GPT-5.6 — divided into the Sol, Terra, and Luna models — occurred less than 24 hours after news broke that the company was spacing out the model's delivery at the request of the Trump administration. What might seem like a mere logistical detail is, in practice, the milestone marking the end of the artificial intelligence wild west.
Until now, the release cycle for frontier models followed a move fast and break things logic. The dynamic was that of an arms race: whoever reached the next level of intelligence first would dictate the market's rules. But the decision to subject GPT-5.6 to government approval on a client-by-client basis inverts this logic. The State ceases to be a reactive regulator chasing the train and becomes the gatekeeper deciding who gets to board. Innovation must now ask for permission.
My point here is that this is not a temporary concession by OpenAI, but the acknowledgment of a new business reality. Giant language models have ceased to be mere software products to become critical infrastructure. By accepting federal scrutiny as a condition for the preview of its models, the company sanctions the idea that frontier AI is too dangerous for the free market. Capitalism ceases to be the arbiter of technology; geopolitics takes the post.
The most profound consequence of this move is the consolidation of the State as the natural gatekeeper of artificial intelligence. We are not just talking about compliance or safety against existential risks. We are talking about discretionary power. If the government can dictate the pace and audience of an AI release under the guise of national security, it holds in its hands the selective key to technological progress. Who receives the model and when ceases to be a commercial decision and becomes a matter of state policy.
This should frighten Silicon Valley evangelists who still dream of libertarian disruption, but it is the inevitable fate of any technology that threatens to rewrite the rules of power. OpenAI is not just launching a smarter model; it is minting a new social contract for innovation. And in this contract, the State is not an obstacle to be bypassed, but the silent partner who signs off at the bottom.
OpenAI spaced out the delivery of the GPT-5.6 models (Sol, Terra, and Luna) to comply with a request from the Trump administration, subjecting the release to government approval on a client-by-client basis.
It marks the end of the 'move fast and break things' era. The State becomes the gatekeeper of frontier AI, meaning innovation must ask for permission and technological progress becomes a matter of state policy rather than a purely commercial decision.
It signifies that large language models are now treated as critical infrastructure rather than mere software. Capitalism is no longer the arbiter of technology; geopolitics and the State dictate the pace and audience of AI releases.