OpenAI forced to beg the government to release GPT-5.6 client by client

OpenAI will not be dropping its latest flagship model, GPT-5.6, for the masses all at once. Instead, the company is implementing a restricted preview phase where every single user must be individually vetted by the U.S. government. Sam Altman confirmed this arrangement in a internal memo, describing it as the fastest possible route to market, despite it being clearly far from his ideal business strategy.

This bottleneck is a direct consequence of the Donald Trump executive order issued on June 2nd, which requires AI labs to grant the government up to 30 days of inspection time for top-tier models. The Department of Commerce, led by Howard Lutnick, has been leaning hard on OpenAI, demanding full sign-offs from various agencies before any public exposure. This follows the recent blocking of Anthropic’s Fable 5 and Mythos 5, which were effectively frozen out of the market due to national security concerns.

The industry is rapidly pivoting to this new reality where the state acts as the primary gatekeeper for frontier models. Following the pattern Anthropic used for its own Mythos model earlier this year, these security reviews are becoming the industry standard. The technical specs of GPT-5.6 are currently secondary to the opaque list of requirements and criteria developers must navigate to earn a stamp of approval.

Welcome to the era of state-managed innovation, where the speed of progress is dictated by the slowest bureaucrat in the room. When the government decides who gets to play with the next generation of intelligence, the dream of decentralized, open-source AI becomes nothing more than a cute marketing slogan for the gullible.

Source: The Information