The Trump administration's emergency export ban on Anthropic's most advanced AI model marks the first time the US has applied such controls to software rather than hardware, after the model reportedly compromised nearly all classified NSA systems within hours.
The Trump administration's emergency export ban on Anthropic's most advanced AI model marks the first time the US has applied such controls to software rather than hardware, after the model reportedly compromised nearly all classified NSA systems within hours.
The Trump administration on June 12 ordered Anthropic to block foreign access to its Fable 5 and Mythos 5 models — the first export controls ever applied to an AI system — after the model penetrated nearly all classified NSA networks within hours during a red-team exercise, according to Senate testimony.
"This tool broke into almost all of our classified systems, not in weeks, but in hours," Senator Mark Warner of Virginia said during a June 11 Banking Committee hearing, citing NSA and US Cyber Command chief General Joshua Rudd.
The directive gave Anthropic 90 minutes to comply, cutting off Five Eyes partners Australia, the UK, Canada and New Zealand without warning. The UK AI Security Institute was locked out of systems it was actively evaluating. Five intelligence agencies from the alliance issued a rare joint statement Monday warning that frontier AI models "are anticipated to exceed current industry expectations" within months, not years.
The ban threatens to upend Anthropic's commercial trajectory. Fable 5 had been publicly available for just three days before the crackdown. Prediction markets price 57% odds of restoration before July 1, but the episode has already triggered a broader reckoning: Europe is watching closely, and former British security minister Tom Tugendhat warned that "every nation will be asking what they need to achieve sovereignty."
A 27-Year-Old Flaw and 271 Firefox Bugs
Mythos, first announced in April, was deemed too capable at finding security vulnerabilities for a public release. Anthropic opened access through Project Glasswing, a controlled program of roughly 200 vetted partners including Amazon, Apple, Google, Microsoft, Nvidia, JPMorgan and the Linux Foundation. The model had already uncovered thousands of real-world vulnerabilities — including a 27-year-old flaw in OpenBSD and 271 new bugs in Mozilla's Firefox 150 browser.
Fable 5, released June 9, uses the same underlying model with safety classifiers designed to intercept potentially dangerous requests. The official trigger for the ban was a jailbreak reported to the Commerce Department by Amazon — a major Anthropic investor and AI competitor. Anthropic said the jailbreak was "narrow and non-universal" and that similar vulnerabilities exist in OpenAI's GPT-5.5, which faces no comparable restrictions.
Industry Pushback and the Path Forward
More than 100 cybersecurity executives from companies including Adobe and Nvidia signed an open letter urging the administration to commit to "an open, scientific and transparent process" for AI risk assessments. The letter acknowledged Mythos is "quite good" at finding software flaws but argued it is "not uniquely good at these tasks," noting that open-source models running 6 to 8 months behind frontier labs can replicate similar capabilities.
The political backdrop complicates the security rationale. The Trump administration in February ordered all federal agencies to stop using Anthropic's models after the company refused contract terms allowing autonomous weapons use and mass domestic surveillance. The Pentagon subsequently designated Anthropic a "supply chain risk," a label the company is challenging in federal court.
Trump told Axios on June 19 that he no longer views Anthropic as a national security threat. "Well, not now, but a week ago, maybe," he said. Anthropic's identity verification policy, set to take effect July 8, could allow the company to restore Fable 5 domestically without requiring the export directive to be fully lifted.
The deeper problem is structural. The US has no consistent framework for evaluating frontier AI models before release. The June 2 executive order asking for voluntary 30-day pre-release access was ignored — Fable 5 launched seven days later with no pre-brief. The ban was, in effect, the mechanism for forcing the cooperation the voluntary framework couldn't compel.
This article is for informational purposes only and does not constitute investment advice.
