OpenAI’s upcoming release of its latest model, GPT-5.6, is expected to mark a significant departure from previous launches. Rather than a broad public rollout, the company will initially distribute the model to a small circle of trusted partners—following direct requests from the Trump administration, as reported by The Information.
During a staff meeting this week, CEO Sam Altman reportedly informed employees that the government would be "approving access customer by customer" throughout a preview phase. Altman added that if the limited release goes smoothly, OpenAI hopes to follow with a broader public release "a couple of weeks later."
In effect, the Trump administration appears to be pressuring OpenAI to adopt a cautious approach similar to what Anthropic has already voluntarily implemented: keeping its most powerful AI models under tight control.
According to The Information, OpenAI’s new model is not only under government review but was developed in close collaboration with federal agencies. The Office of the National Cyber Director and the Office of Science and Technology Policy reportedly requested the limited release.
The Trump administration—which initially championed a "hands-off" approach to AI—has increasingly pushed for federal oversight of cutting-edge models. Earlier this month, Trump signed an executive order directing certain AI companies to voluntarily submit new models for government testing and evaluation before public release.
Earlier in 2026, Anthropic stirred controversy when it announced that its frontier cyber model, Claude Mythos, would only be released to a select group of partners through a program called Project Glasswing. Anthropic justified the decision by arguing that the model’s capabilities could, in the wrong hands, cause more harm than good. Since then, observers have debated whether Anthropic’s stance is a genuine safety measure or a marketing ploy—with the truth likely lying somewhere in between.
Cybercriminals have long used automated tools, but in the era of generative AI, they now have even more powerful digital weapons. Large language models (LLMs) have proven adept at writing malware, and some can even execute entire ransomware attacks autonomously.
The core concern with frontier cyber tools like Mythos is their ability 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 threat to any organization running complex software infrastructure. However, because these models remain closed to the public, it is difficult to assess their true risk.
via TechCrunch AI