For the first time in the company’s history, OpenAI has staggered the release of a flagship AI model series at the explicit request of the United States government — a development that signals a profound and largely unannounced shift in how Washington intends to control access to cutting-edge artificial intelligence. The move raises fundamental questions about who ultimately decides which companies, developers, and users gain access to the most powerful AI tools being built today.

What Happened: GPT-5.6 Released Only to ‘Trusted Partners’

OpenAI announced on June 26, 2026, that it had begun “a limited preview of the GPT 5.6 series… for a small group of trusted partners whose participation has been shared with the government.” The GPT-5.6 lineup consists of three models: Sol, described as the flagship; Terra, positioned as a balanced model for everyday use; and Luna, a faster, lower-cost option.

The company framed the restricted rollout as a “short-term” arrangement intended to bridge the gap while the Trump administration develops a formal framework for reviewing frontier AI models. That framework stems from a June 2 executive order signed by President Donald Trump — an order whose specific review mechanisms have not yet been fully defined or made public.

OpenAI stated that it believes in “broad access” and that it plans to make the new models generally available in the coming weeks. But the company also acknowledged that the current arrangement keeps its most advanced tools out of the hands of ordinary users, developers, and enterprises while government vetting processes are conducted behind closed doors.

A Pattern Emerges: Anthropic Was First

OpenAI is not the first major AI developer to comply with such a government directive. Earlier in June 2026, Anthropic disabled access to its Fable 5 and Mythos 5 models just three days after launch, following a US government export control directive that cited national security concerns. According to reporting by Axios and The Verge, Mythos 5 has since been partially restored after Anthropic addressed the government’s concerns, while restrictions on Fable 5 remain in place.

US Commerce Secretary Howard Lutnick addressed the Anthropic situation in a June 26 letter to Anthropic co-founder Tom Brown, stating that the company’s engagement with authorities had “yielded significant progress.” The letter confirmed that direct, bilateral communications between government officials and AI executives are now a functional part of how frontier model releases are being managed.

Together, these two cases — Anthropic’s forced pause and OpenAI’s pre-emptive restriction — represent the emergence of an informal but effective government gatekeeping mechanism for frontier AI, one that is operating without a clear statutory basis or public oversight process.

The Regulatory Reversal: From ‘Beautiful Newborn Baby’ to Controlled Asset

The current posture represents a notable reversal from the Trump administration’s original stance on AI. After returning to office in 2025, Trump positioned himself as a champion of minimal AI regulation, famously describing AI as “a beautiful newborn baby” central to competing with China, and warning against “politics or stupid rules” that could hinder its growth.

In recent weeks, however, the same administration has moved toward closer oversight of frontier AI models, citing growing cybersecurity and national security concerns as increasingly capable systems become available. Critics and policy commentators have noted that this shift is occurring without a clear legal framework — meaning that compliance by companies like OpenAI and Anthropic is currently being secured through informal government requests rather than enforceable law.

Some mainstream media and policy commentators have argued that the administration is expanding its influence over the rollout of cutting-edge AI systems in a manner that reduces predictability for developers and could ultimately weaken the competitiveness of US AI firms in the global market.

OpenAI’s Own Warning: ‘This Shouldn’t Be the Norm’

OpenAI itself appeared to push back, at least rhetorically, against the arrangement it agreed to comply with. According to TechCrunch’s reporting, OpenAI stated: “We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.”

That statement is significant. OpenAI is simultaneously complying with a government request and publicly signaling its concern that such requests could become institutionalized. The company’s position illustrates the bind that major AI developers now find themselves in: powerful enough to shape global technology, but operating under increasing pressure from the government that hosts them and the executive orders that could define their future operating environment.

The Broader Context: Government Influence Over AI Is Not New

This episode does not emerge from a vacuum. In March 2025, the House Committee on the Judiciary sent a letter to OpenAI CEO Sam Altman requesting documents about the nature and extent of the company’s interactions with the executive branch regarding content moderation of its AI models. The committee, which had previously investigated what it characterized as the Biden-Harris administration’s efforts to pressure social media platforms and AI companies to moderate content in line with government preferences, sought to understand whether similar dynamics had shaped OpenAI’s model development.

That earlier committee investigation documented how, between July 2023 and July 2024, sixteen AI companies — including OpenAI — pledged to abide by Biden-era White House “voluntary commitments” on AI to mitigate alleged “harmful bias” and “algorithmic discrimination.” The committee’s report characterized these arrangements as a form of regulatory pressure that blurred the line between voluntary industry standards and government-directed content control.

Now, under a different administration with a different ideological framework, the mechanism has shifted — from content-focused pressure to access-focused gatekeeping — but the underlying dynamic remains: the executive branch is shaping what AI models do and who can use them, outside of any formal legislative process.

Who Decides Who Gets Access?

Perhaps the most consequential question raised by these developments is one that neither OpenAI, Anthropic, nor the administration has fully answered publicly: what criteria determine which partners are “trusted” enough to receive early access to frontier AI models under government-vetted arrangements?

The current process, as described, involves companies identifying a small group of partners, sharing that list with the government, and proceeding on that basis while a formal review framework is developed. That framework — rooted in Trump’s June 2 executive order — has not yet been publicly detailed in terms of its specific criteria, the agencies involved, or the appeals process available to companies denied access.

What is documented is the outcome: two of the most powerful AI developers in the world have now delayed public access to their most capable models at Washington’s request, and the government is actively corresponding with AI executives about which of their products can be released and under what conditions.

Conclusion: A New Architecture of AI Control

The restriction of GPT-5.6’s rollout is not, on its face, a dramatic event. OpenAI has promised general availability within weeks, and the company’s stated cooperation is framed as a temporary bridge measure. But the precedent being established is worth examining carefully.

For the first time, a US presidential executive order has triggered the pre-emptive restriction of a flagship AI model release. Two major AI companies have now complied with government directives to limit public access to frontier models. A Commerce Secretary is corresponding directly with AI co-founders about whether their products have adequately addressed government concerns. And all of this is unfolding without a clear statutory framework, without public criteria for what triggers a restriction, and without a formal mechanism for independent review.

Whether the emerging architecture of government oversight over frontier AI ultimately serves the public interest — in security, in competition, or in access — will depend entirely on the transparency and accountability of processes that, at present, remain largely shielded from public view.

This article draws on reporting from RT World News, TechCrunch, The Washington Post, and the House Committee on the Judiciary.