OpenAI’s upcoming release of its newest model, GPT 5.6, is set to break from tradition. Rather than making it available to the general public, the company intends to restrict access to a small circle of trusted partners—a move reportedly prompted by directives from the Trump administration. According to inside reports, CEO Sam Altman informed staff during a meeting this week that the government would be "approving access customer by customer" throughout a preview phase. Altman also indicated that if this limited rollout goes smoothly, OpenAI hopes to follow up with a broader public release "a couple of weeks later."

In essence, the Trump administration appears to be pushing OpenAI to adopt a strategy similar to what Anthropic is already doing voluntarily: keeping its most advanced AI models out of public hands. The Information reports that OpenAI’s new model is not only under government review but that company staffers "worked closely" with officials on the release plan. The agencies specifically requesting the restricted rollout include the Office of the National Cyber Director and the Office of Science and Technology Policy. Although the Trump administration initially positioned itself as favoring a "hands-off" approach to AI, it has recently moved toward federal oversight of new models. Earlier this month, Trump signed an executive order requiring certain AI companies to voluntarily submit new models to the government for testing and evaluation before public release.

Earlier this year, Anthropic sparked considerable controversy by announcing that its new frontier cyber model, Claude Mythos, would only be shared with a select group of partners through a program called Project Glasswing. Anthropic argued that the model was simply too powerful and, in the wrong hands, could cause more harm than good. Since then, observers have debated whether Anthropic's stance is a genuine attempt to prevent misuse or merely a marketing gimmick. The truth likely lies somewhere in between. Cybercriminals have long used automated tools, but in the era of generative AI, they now have even more digital firepower. Large language models have proven capable of writing malware, and some can even execute entire ransomware attacks autonomously. The specific concern with frontier cyber tools like Mythos is their potential to identify and exploit software vulnerabilities at speeds no human analyst can match. Given that many software systems contain hidden bugs that serve as entry points into enterprise networks, this poses a significant problem for any organization running complex software infrastructure. However, since these models remain closed to the public, it's difficult to assess just how serious the threat truly is.