Invest2 publishers3 min readPublished
Commerce's export directive took Anthropic's two newest models offline three days after launch
Anthropic says a letter from Commerce Secretary Howard Lutnick barred every foreign national from Fable 5 and Mythos 5, inside the United States as well as outside, and the only way to comply was to switch both models off for all customers.
The Investor · Invest desk
What happened
- Anthropic said the government issued an export control directive suspending all access to Fable 5 and Mythos 5 by any foreign national, inside or outside the United States, including its own foreign-national employees.
- To comply, the company disabled both models for every customer rather than only foreign nationals, and Claude's landing page said on Friday evening that Fable 5 was temporarily unavailable.
- The letter came from Commerce Secretary Howard Lutnick and was written with help from Bureau of Industry and Security officials, according to an administration official cited by NBC News.
- Anthropic said the government supplied only verbal evidence of the safeguard workaround and that the same weakness was likely present in AI systems built by other companies.
- The two models share a technical foundation: Fable 5 went to the public with heavy cybersecurity and biology restrictions, while Mythos 5 went without those safeguards to selected partners.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- precedent A deployed frontier model has now been switched off by federal instruction. Every subsequent release carries an availability term that no vendor service agreement controls.
- exposure Anthropic says the jailbreak probably exists in rival systems. A customer spreading spend across two or three model vendors is not obviously spreading this particular risk.
- decision The June 2 executive order asked agencies to build a voluntary early-access route to frontier models; vendors now have to judge whether volunteering sooner buys any protection against a directive like Friday's.
- contradiction The Financial Times reported last week that the NSA was using Mythos for offensive cyberattacks. One arm of the government was operating the model another arm has now ordered suspended.
Anthropic had Fable 5 and Mythos 5 in the market for three days before the suspension [1]. Its launch note was cautious about the same domain the government raised: "Releasing a model this capable comes with risks. Without safeguards, Fable 5's capabilities in areas like cybersecurity could be misused to cause serious damage," the company said on Tuesday [12]. By Friday it was disputing the remedy: "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" [10].
The technical case is not public. Anthropic said the government identified a specific technique that could let users circumvent some narrow safeguards on Fable 5 [8], and that it was given only "verbal evidence" of the problem [9]. NBC News reported that it was unclear on Friday night why the Commerce Department was specifically worried about foreign nationals exploiting the weakness [21].
The narrow reading is Anthropic's own: the company said it believed the directive stemmed from "a misunderstanding" and hoped to reinstate access as soon as possible [14]. Then there is the history. In February, President Donald Trump and Defense Secretary Pete Hegseth moved to bar Anthropic's products from federal agencies after the company sought stronger guardrails on how the Pentagon used its technology [15]; Trump posted that "The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War" [16]. Anthropic sued on free-speech grounds, won before a federal judge in California, and the case is still live in a Washington federal court [17]. On that reading, Friday is one company's politics.
The third reading is the one a buyer of frontier capacity has to price. An export rule scoped to persons, covering foreign nationals inside the United States and Anthropic's own foreign-national staff [2], cannot be enforced selectively on a hosted endpoint. That leaves a shutdown for everyone. I think this reading survives, and the reason is procedural: the trigger was a verbal finding, the counterparty is a cabinet department, and reinstatement sits with that department. Model deprecation and rate limits were always in the risk register; a nationality-scoped suspension of a shipped model was not, and NBC reported this appears to be the first time a leading AI company has pulled a publicly deployed model offline because of federal intervention [5].
Two developments would undercut that. A quick restoration accompanied by a published statutory basis would make this an administrative episode with a three-day cost. A comparable instruction to another model vendor would make it a finding about a class of systems, which is what Anthropic itself claims, saying the workaround was likely also present in other companies' models [9]. Anthropic said in the same statement, "We believe the government should have the ability to block unsafe deployments" [20].
What to watch
- Whether Fable 5 and Mythos 5 come back, and whether Commerce publishes the statutory authority behind the letter.
- Whether the Bureau of Industry and Security sends a comparable directive to another frontier model vendor.
- Whether enterprise customers of Fable 5 receive service credits or disclose the outage in their own filings.