The public debut of GPT 5.6, OpenAI's latest frontier model, will not follow the company's traditional release pattern. As reported by TechCrunch, the Trump administration has requested a "slow roll" of the technology, limiting access to a select group of vetted partners.

Government Oversight in Real-Time

CEO Sam Altman reportedly informed employees that the government will be "approving access customer by customer" during a preliminary preview period. A broader general release is expected only after this restricted phase proves successful, likely within a few weeks.

The push comes from the Office of the National Cyber Director and the Office of Science and Technology Policy. This represents a significant shift toward federal oversight, reinforced by a recent executive order requiring AI companies to voluntarily submit frontier models for government testing before public availability.

The Shadow of Claude Mythos

This restricted approach mirrors the strategy previously employed by Anthropic for its Claude Mythos model. Through Project Glasswing, Anthropic limited access to a small circle of partners, arguing that the model's capabilities were too potent and could cause systemic harm if misused.

The core concern is the ability of these advanced models to identify and exploit software vulnerabilities at speeds that far exceed human capabilities. With LLMs already demonstrating proficiency in writing malware and executing autonomous ransomware attacks, the U.S. government views GPT 5.6 as a potential cybersecurity liability that requires strict vetting.

A New Era of AI Regulation

The intervention highlights the growing friction between frontier labs and Washington. While OpenAI has previously focused on defensive AI initiatives to automate patching, the White House is now moving toward a model of direct intervention in product launch calendars to mitigate catastrophic risks.