This is a significant moment for the AI industry and I have been watching it closely. The US Department of Commerce has officially approved OpenAI to release its advanced AI model GPT-5.6 on a large scale, ending a period of government-imposed restrictions that had limited access to a small group of approved entities.
OpenAI confirmed that GPT-5.6 Sol is expected to go public this Thursday alongside two other models named Terra and Luna, pending final preparations. That is a three-model release in a single day, which is not a small moment for anyone following where AI development is headed right now.
The background to this approval matters. The US government had required OpenAI to follow a phased release strategy for GPT-5.6 over national security concerns. Last month the model was only accessible to a limited number of government-approved organisations, a restriction OpenAI made clear it was not happy with at the time. The company stated publicly that phased release was not its preferred approach, which made the government review process feel like an obstacle rather than a partnership.
The testing that led to this approval was conducted by the AI Standards and Innovation Center under the US Department of Commerce. During the evaluation period OpenAI deployed a team of technical experts to Washington to respond directly to questions from government officials as they arose. That level of on-the-ground coordination suggests the approval process involved genuine back and forth rather than a simple paperwork review.
With the restrictions now lifted, Thursday's release marks the formal end of that temporary control measure. The timing also lines up with a new US executive order on AI that is reportedly close to being released. The order is expected to establish a formal evaluation framework for advanced AI model releases going forward, essentially turning what happened with GPT-5.6 into a structured process that future models from any company will go through before public deployment.
OpenAI has stated that AI companies and the US government are already coordinating within the existing framework, which suggests the industry is not waiting for the executive order to land before aligning with the direction it is heading.
This situation is not unique to OpenAI. A similar phased release approach was previously applied to Anthropic for its Mythos and Fable models, which means the government has been developing this evaluation approach across multiple frontier AI companies rather than singling out any one organisation. The pattern suggests that advanced AI model releases above a certain capability threshold are now entering a category where government review before public deployment is becoming standard practice rather than an exception.
Thursday's full release of Sol, Terra, and Luna will be worth watching closely.
