White House Asks OpenAI to Slow-Roll GPT-5.6 Release Over Safety Concerns
White House Intervenes in GPT-5.6 Launch, Marking Historic First in U.S. AI Oversight
The Trump administration has asked OpenAI to restrict the release of its next flagship model, GPT-5.6, to a small set of government-approved partners before any broader public rollout — marking the first time the U.S. government has preemptively asked an American AI company to limit the launch of a model prior to its release, according to Axios reporting from June 25, 2026. The request came from two White House offices: the Office of the National Cyber Director and the Office of Science and Technology Policy, both citing security concerns about the model's capabilities.
OpenAI CEO Sam Altman disclosed the limited rollout plan in a memo to employees, stating the government would be "approving access customer by customer during this preview period," according to The Information, first reported and subsequently cited by CNN and Axios. Altman reportedly told employees the staggered approach was the fastest path to a broad release, and that he hoped a wider rollout would follow "a couple of weeks later" if the review went well, according to SiliconAngle.
OpenAI's own position on the arrangement was measured. In a statement reported by Axios, the company said: "We've made clear to the U.S. government that this is not our preferred long term model, and will work with them and others in industry to achieve a more sustainable approach for future releases."
Why GPT-5.6 Triggered Government Scrutiny
The administration's concern centers on GPT-5.6's advanced capabilities. According to Axios, a source said the government intervened because the model has "Mythos-like" capability — a reference to Anthropic's Mythos model, which was itself subject to a sweeping export control action by the U.S. Commerce Department less than two weeks earlier. CNN reported that both OpenAI and the administration view GPT-5.6 as "on par" with Anthropic's Mythos model.
The cybersecurity dimension of these models is not abstract. According to Axios reporting from June 23, 2026, OpenAI's GPT-5.5-Cyber achieved an 85.6% score on CyberGym, an internal benchmark measuring whether an AI agent can reproduce known software vulnerabilities. Anthropic's Mythos 5 scored 83.8% on the same evaluation. These figures illustrate why frontier AI capabilities have become a national security flashpoint: models that can identify and replicate software vulnerabilities at scale represent a potential tool for attacking critical infrastructure.
GPT-5.6 is internally codenamed "kindle-alpha," according to Cryptobriefing. Following news of the government-requested slow rollout, the Polymarket prediction market showed the probability of GPT-5.6 being released by June 30, 2026 had dropped to 7.7%, down from 10% just 24 hours prior.
Altman discussed GPT-5.6 directly with Commerce Secretary Howard Lutnick on June 24, 2026, with Lutnick reportedly wanting to ensure all relevant parts of the government had tested and approved the model before any broader release, according to Axios.
The Anthropic Precedent: A More Forceful Intervention
The GPT-5.6 situation cannot be understood in isolation. It follows a more aggressive federal action taken against Anthropic just two weeks earlier. On June 12, 2026, the U.S. Commerce Department's Bureau of Industry and Security (BIS) issued an emergency export control directive ordering Anthropic to immediately suspend access to its Fable 5 and Mythos 5 models for all foreign nationals — including Anthropic's own non-citizen employees — citing authority under the Export Control Reform Act of 2018, according to multiple legal analyses from CSIS and A&O Shearman.
In an official statement posted to its website and on X, Anthropic confirmed: "The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." The directive was received at 5:21 PM ET on June 12, 2026. Because Anthropic could not reliably screen users by nationality in real time, it disabled both models entirely for all customers worldwide to ensure compliance, while all other Anthropic models remained unaffected.
Anthropic pushed back sharply on the action's rationale. In its official statement, the company said: "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people." It added: "If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."
Notably, the intervention against Anthropic came from the Commerce Department, while the request to OpenAI came from the White House. According to The Verge, as cited by multiple outlets, the GPT-5.6 arrangement is more permissive than the restrictions applied to Anthropic's Mythos 5 and Fable 5, which were subject to limitations on foreign access. The inconsistency in both the source of authority and the severity of intervention across the two companies has drawn scrutiny from legal experts and industry observers alike.
For context, Anthropic had initially restricted access to Mythos to approximately 50 companies and organizations managing critical infrastructure as part of its Project Glasswing, according to The Conversation.
The Regulatory Backdrop: Trump's AI Executive Order
The broader regulatory context was set on June 2, 2026, when President Trump signed an executive order titled "Promoting Advanced Artificial Intelligence Innovation and Security." According to NPR's reporting, the order asks AI companies to voluntarily submit their most powerful models for government testing up to 30 days before releasing them to the public. It also directs multiple federal agencies to build a voluntary testing framework, establishes a classified AI benchmarking process to be overseen by the NSA Director, and creates an AI cybersecurity clearinghouse.
An earlier draft of the order had proposed a 90-day pre-release review window, according to CFR reporting from June 2026, but that was cut to 30 days amid concerns that a longer window would hurt U.S. AI competitiveness relative to China. Crucially, the framework is voluntary — it does not impose mandatory preclearance requirements on AI developers. The government's preemptive request to OpenAI regarding GPT-5.6 therefore exists in a legal gray zone: it is not a statutory mandate, but rather a request backed by the implied weight of federal authority and, in Anthropic's case, the demonstrated willingness to use emergency export control powers.
An Ad Hoc, Opaque Approach: Expert Reactions
The emerging federal posture toward frontier AI has drawn pointed criticism for its inconsistency and lack of transparent legal grounding. Brad Carson, head of Public First, a bipartisan pro-AI safety super PAC, offered a direct assessment to CNN: "Right now, you have an ad hoc, personalized, opaque, possibly lawless approach."
Carson acknowledged the legitimacy of the underlying concern while challenging the method of intervention: "It is certainly appropriate for the government to recall dangerous products, including AI models, but it has to be done in a way consistent with transparency and basic fairness."
These critiques point to a structural problem: there is currently no transparent, statutory framework governing when and how the federal government can intervene in the release of a commercial AI model. The actions taken against Anthropic and the request made to OpenAI emerged from different agencies, under different legal theories, with different levels of coercive force. Axios confirmed that the OpenAI request represents the first time the U.S. government has preemptively asked an American AI company to restrict the launch of a model before release — a significant policy milestone that arrived without a clear legal foundation or public deliberation.
What This Means for the AI Industry — and for Users
For the AI industry, the implications are significant and unresolved. The federal government has now demonstrated both the willingness and the claimed authority to intervene in AI model releases at multiple levels — from export controls imposed by the Commerce Department to direct White House requests for staggered rollouts. Whether these interventions will be applied consistently, and under what legal standard, remains unclear.
The voluntary framework established by Trump's June 2 executive order creates a pathway for pre-release government testing, but it does not bind AI companies in the way that the BIS export control directive bound Anthropic. The GPT-5.6 arrangement — a customer-by-customer government approval process during a preview period — is a novel mechanism with no established precedent in U.S. technology regulation.
For businesses and individual users who rely on frontier AI tools, the practical upshot is uncertainty. Access to the most capable models may now be subject to government review processes that are opaque and variable. OpenAI has signaled it is cooperating with the current arrangement while seeking a more durable long-term framework. Whether that framework materializes — and whether it applies equally across the industry — will define the regulatory landscape for frontier AI in the United States for years to come.
For more tech news, visit our news section.
What to Watch Next
Several developments are worth tracking in the coming weeks. First, whether OpenAI's GPT-5.6 clears the government review process and moves to a broader public release within the timeline Altman described to employees. Second, whether the Commerce Department takes any further action under the export control framework established by its intervention against Anthropic — particularly given that Anthropic challenged the rationale behind that action in its public statement. Third, whether the voluntary pre-release testing framework established by Trump's June executive order evolves into something more formal and mandatory, and whether the proposed 30-day review window becomes an industry norm or a floor for more restrictive demands.
The fact that the government's first preemptive request to limit a model release came through an informal White House ask — rather than through any statutory or regulatory mechanism — underscores how unsettled this policy terrain remains. The next few weeks may offer the first real test of whether the administration's approach to frontier AI oversight stabilizes into a coherent framework or continues to develop on a case-by-case basis.
The rapid pace of AI development has direct implications for how we work, make decisions, and manage our cognitive load. As governments grapple with the risks of increasingly powerful AI tools, individuals and organizations face their own challenge: staying informed and using these technologies wisely. Moccet is built to help you do exactly that — cutting through the noise to deliver the tech and health insights that matter for your productivity and wellbeing. Join the Moccet waitlist to stay ahead of the curve.