President Trump signed Executive Order 14409 on June 2, 2026, directing federal agencies to create an opt-in, pre-release review process that gives the government early visibility into advanced AI systems before they reach the public. The order, titled "Promoting Advanced Artificial Intelligence Innovation and Security," is organized around three core components: upgrading federal and critical infrastructure cybersecurity, establishing a voluntary framework for pre-release engagement on frontier AI models, and prioritizing criminal enforcement against AI-enabled cybercrimes. Within 60 days, Treasury, NSA, CISA, NIST, and other federal officials were directed to develop a classified benchmarking process to assess the advanced cyber capabilities of AI models and determine when a model should be designated a "covered frontier model." Developers would be able to provide the federal government with access to covered frontier models, subject to confidentiality and intellectual property protections, for up to 30 days before releasing those models to other trusted partners.
The first and sharpest test of the new order was Anthropic. The U.S. government ordered Anthropic to take down its new AI models, Claude Fable 5 and Mythos 5, just three days after their public launch on June 12, citing national security and export control powers; the Commerce Department mandated the suspension of access for all foreign nationals, leading to a global shutdown, following reports of a jailbreak bypassing safety guardrails. The order required Anthropic to halt access for any foreign national, including the company's own non-citizen employees, regardless of location; because Anthropic could not selectively filter users by nationality across dozens of global cloud platforms in real time, it enforced a universal shutdown. On June 30, the Commerce Department withdrew the export-control directive, and on July 1, Anthropic redeployed Fable 5 and Mythos 5 globally. OpenAI's experience was less dramatic but structurally similar. The Trump administration asked OpenAI to restrict the launch of GPT-5.6 to a small group of government-approved partners, marking the first time the White House directly shaped the rollout schedule of a new AI model before it reached the public; the federal government individually vetted each organization that received access during the initial preview window. The Commerce Department's Center for AI Standards and Innovation carried out the evaluation, and OpenAI dispatched technical staff to Washington to respond to the agency's inquiries. OpenAI released GPT-5.6 publicly on July 9, 2026, as three tiers named Sol, Terra, and Luna, after a U.S. Commerce Department review cleared the models for broad access.
The two episodes together established a new operational reality. What they produced is the clearest precedent yet: in the United States in 2026, releasing a frontier AI model now runs through Washington first, whether the law says so or not. Executive Order 14409 directed the NSA, CISA, the Treasury Department, and the White House's National Cyber Director to build a classified benchmarking process to determine which AI systems qualify as covered frontier models, and a voluntary framework governing how those models are reviewed before release. A draft of that framework has already circulated to OpenAI, Anthropic, and Google, with the NSA and CAISI named as reviewers. The executive order gave the government 60 days to design the voluntary process; that clock runs out on August 1.
Skeptics argue the architecture is more coercive than its label suggests. Dean Ball, a former White House AI adviser, argues that the ostensibly voluntary mechanism has evolved into a de facto involuntary licensing regime, leading to heavy-handed government restrictions on model releases. A White House official pushed back, telling CNBC that the administration did not give OpenAI a "green light, approval, or clearance" and that such permission "is not required or granted," adding that decisions on timing and scope of releases "rest entirely with the companies." The practical reality, critics note, is that frontier AI model availability is now a policy variable, not just a commercial one; the suspension of Fable 5 and the gating of GPT-5.6 established a new risk category in which a commercial AI service can be pulled globally, instantly, without prior statutory process and without an appeals mechanism. Some smaller AI companies also worry that a definition tuned to leading closed models such as Mythos and GPT-5.6 could exclude open-source models from any meaningful representation in the process.
The August 1 deadline for the formal framework means that, within days, Washington will either publish rules that give the AI industry some predictability, or leave the biggest labs navigating an ad-hoc veto power with no written standards at all.