The government is now deciding who gets to use American AI first
In a quiet but consequential shift, the U.S. government has stepped between a finished technology and the world waiting to receive it. OpenAI's GPT-5.6 sits complete yet unreleased, held at a federal threshold while the Trump administration conducts a cybersecurity review and decides who may access it first. Both OpenAI and Anthropic have accepted this arrangement without public resistance, marking a moment when the state's hand became visible in the chain that connects AI laboratories to the rest of humanity.
- GPT-5.6 is finished and functional, yet locked behind a federal checkpoint with no announced timeline for public release.
- The Trump administration has claimed the role of gatekeeper, granting early access only to explicitly approved customers while cybersecurity reviews proceed.
- Anthropic has mirrored OpenAI's compliance, suggesting this is not an isolated concession but a coordinated shift across the frontier AI industry.
- Neither company has publicly pushed back, raising the question of whether cooperation reflects genuine agreement, strategic calculation, or both.
- A precedent is quietly hardening: if this model holds, government oversight of frontier AI distribution may become the rule rather than the exception.
OpenAI has postponed the public release of GPT-5.6 at the Trump administration's request, granting federal officials a window of exclusive early access before the model reaches the broader market. The stated rationale is cybersecurity — government reviewers want time to examine the system for vulnerabilities and dangerous capabilities before it spreads widely. Anthropic has adopted the same posture with its own new models, limiting distribution in parallel.
What distinguishes this moment is not the delay itself but the mechanism behind it. The U.S. government is now directly deciding which organizations can access cutting-edge American AI systems before anyone else — a function that companies like OpenAI previously exercised on their own. The security concerns are real enough; frontier models do raise genuine questions about misuse by hostile actors or criminal organizations. But the arrangement also establishes something larger: a precedent in which the state positions itself as arbiter of who touches the most powerful AI tools first.
Neither company has resisted publicly, suggesting either that they find the review reasonable or that they regard cooperation with federal authorities as a necessary long-term investment, even at the cost of short-term revenue. For the organizations and users waiting to deploy GPT-5.6, the uncertainty is concrete — the model exists, it works, and yet it remains inaccessible. What happens next will be decided not in a laboratory, but in a government office.
OpenAI has postponed the public launch of GPT-5.6, its latest flagship model, at the request of the Trump administration, which is seeking exclusive early access to the system before it becomes available to the broader market. The delay marks a significant moment in how advanced artificial intelligence technology moves from laboratory to the world—one in which the U.S. government has inserted itself as a gatekeeper, at least temporarily.
The administration's request centers on cybersecurity concerns. Before OpenAI and Anthropic, another leading AI company, release their newest models to the general public, federal officials want time to evaluate the systems for potential security vulnerabilities and risks. This review period has effectively created a window during which access to GPT-5.6 will be restricted to customers who have received explicit approval from the Trump administration. Anthropic has adopted the same approach with its own new models, limiting distribution in parallel with OpenAI.
What makes this development noteworthy is not merely the delay itself—product launches are postponed all the time for various reasons—but rather the mechanism driving it. The U.S. government is now exercising direct control over the distribution of American frontier AI technology, deciding which organizations and individuals can use cutting-edge systems before anyone else. This represents a shift in how the relationship between government and private AI companies operates. Previously, companies like OpenAI made these decisions independently, releasing products on their own timelines and to their own chosen audiences.
The security review rationale is straightforward enough. Advanced AI models, particularly those approaching or exceeding human-level capabilities in certain domains, do present genuine security questions. Governments worldwide have grown increasingly concerned about the potential misuse of frontier AI—whether by hostile foreign actors, criminal organizations, or bad actors within their own borders. A period of controlled access, during which federal cybersecurity experts can examine the model for weaknesses or dangerous capabilities, offers a way to mitigate some of these risks before the technology spreads widely.
Yet the arrangement also signals something broader about the emerging relationship between the state and AI development. By positioning itself as the arbiter of who gets access to GPT-5.6 first, the federal government is establishing a precedent. If this model holds—if OpenAI and Anthropic continue to accept such restrictions, and if other AI companies follow suit—then government oversight of frontier AI distribution could become standard practice rather than an exception. The question of who controls access to the most powerful AI systems is, in many ways, a question about power itself.
Neither OpenAI nor Anthropic has publicly resisted the arrangement, suggesting either that they view the security review as reasonable or that they see cooperation with federal authorities as strategically necessary. The companies operate in a regulatory environment that is still taking shape, and maintaining good relations with government agencies may feel like a prudent long-term investment, even if it means delaying revenue and market share in the short term.
For users and organizations waiting to deploy GPT-5.6, the delay introduces uncertainty. No public timeline has been announced for when the cybersecurity review will conclude or when broader access will be granted. The model exists, it works, but it remains locked behind a federal checkpoint. What happens next depends on decisions being made in government offices, not in OpenAI's labs.
La Conversación del Hearth Otra perspectiva de la historia
So the government asked OpenAI to hold back a finished product. Did OpenAI push back at all?
Not publicly. They seem to have accepted it as a condition of doing business in this environment.
Why would they agree to that? They're losing money by not selling it.
Short-term revenue versus long-term regulatory goodwill. If you're building the most powerful AI systems in the world, you need the government on your side.
But this sets a precedent, doesn't it? Once you let them do it once...
Exactly. It becomes normal. The next model, the one after that. Eventually it's just how things work.
And nobody else can use it in the meantime?
Only people the Trump administration approves. It's a closed circle while the review happens.
How long does a cybersecurity review take?
That's the thing—nobody's said. It could be weeks or months. The uncertainty is part of the pressure.