In an unprecedented move, the Trump administration has ordered Anthropic to disable its most advanced artificial intelligence models — Fable 5 and Mythos 5 — for all users worldwide, citing national security concerns triggered by a surprisingly simple jailbreak: the three-word prompt "fix this code."
The export control directive, issued by the Commerce Department on Friday, June 12, 2026, forces Anthropic to suspend access to both models for any foreign national, regardless of whether they are inside or outside the United States. Since U.S. export control rules classify sharing controlled technology with non-citizens as an export — even within company walls — Anthropic had no choice but to shut down the models entirely for everyone.
How Three Words Sparked a Government Shutdown
According to a detailed report by cybersecurity expert Katie Moussouris, founder of Luta Security and a former Microsoft security engineer, the vulnerability works in a deceptively straightforward way. Researchers at Amazon discovered that by presenting Fable 5 with code containing known, unpatched vulnerabilities and then asking it to "fix this code," the model would generate working patches. While this sounds like a defensive capability, the same technique could be weaponized by attackers to automatically scan codebases for exploitable flaws.
Amazon CEO Andy Jassy personally escalated the issue to the White House, which prompted the Commerce Department to act. Anthropic said it received the order at 1:00 PM ET on Friday with no advance warning, followed by a formal written directive around 5:30 PM. The company had only publicly unveiled the models three days earlier, on June 9.
Cybersecurity Experts Push Back
The decision has drawn sharp criticism from the cybersecurity community. An open letter organized by Alex Stamos, former Chief Security Officer of Meta and Facebook, has gathered roughly 100 signatories from companies including Nvidia, Adobe, Zoom, Google, and Sophos. The letter, hosted at freefable.org, argues that the export controls are counterproductive.
"To pull the best capabilities away from defenders without a good reason when our adversaries are rapidly advancing is dangerous," the letter warns, pointing to rapid progress in Chinese AI models. The signatories argue that the "fix this code" capability is not unique to Fable — OpenAI's GPT-5.5, Anthropic's own Claude Opus and Sonnet, and even open-source models can perform equivalent code security reviews.
Moussouris drew a historical parallel to the 1990s fight over encryption export controls, when cryptographer Adam Back printed RSA code on T-shirts as a protest. She suggested critics of the Fable ban print shirts reading "fix this code" on one side and "this shirt is a munition" on the other — a reference to the U.S. government's past classification of encryption software as a munition.
Mythos: The Model That Could Hack Autonomously
Mythos 5, the underlying base model, represents a significant leap in AI capability. It was the first AI to pass both cybersecurity test ranges operated by the U.K. AI Security Institute, and can autonomously identify and chain multiple vulnerabilities to orchestrate end-to-end cyberattacks. Fable 5, a fine-tuned version with aggressive safety guardrails, was designed specifically for defensive use.
Anthropic has defended its models, arguing that the jailbreak concern is overblown. "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," the company said in a statement. "If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."
The company also noted that it had pre-tested the models with U.S. government agencies and received formal approval before launch — making the sudden reversal particularly baffling for industry observers.
Anthropic's Strained Relationship With Washington
The Fable-Mythos ban is the latest chapter in Anthropic's increasingly rocky relationship with the federal government. Earlier this year, the Department of Defense labeled Anthropic a "supply chain risk," banning defense contractors from using its technology. Defense Secretary Pete Hegseth publicly defended the blacklist on social media, arguing that "every passing day" proves the decision was justified.
Anthropic has filed a lawsuit to challenge the DOD designation, and senior company representatives met with Trump administration officials in Washington, D.C. on Monday, June 15, to try to resolve the dispute. As of publication, access to Fable 5 and Mythos 5 remains suspended for all users.
Broader Implications for AI Regulation
The incident raises serious questions about how the U.S. government approaches AI safety and export controls. Council on Foreign Relations analysts described the decision as reflecting "an administration trying to sustain its deregulatory, innovation-first posture while confronting the novel cyber risks posed by powerful new tools."
The move has also frustrated U.S. allies. Canada, the United Kingdom, and European Union member states had been relying on Mythos through Project Glasswing, an initiative to identify cybersecurity vulnerabilities in critical infrastructure. The blanket ban cuts them off too.
The Fable 5 and Mythos 5 situation may become a defining moment for how the U.S. regulates frontier AI — and whether the government's approach to export controls is equipped for a world where the most powerful AI models can be triggered by a simple three-word phrase.
Top comments (0)