Days after unveiling what it called its most capable artificial-intelligence systems yet, Anthropic has abruptly switched them off. The company says it disabled its two newest models — Claude Fable 5 and Claude Mythos 5 — to comply with a US government export-control directive, delivered late on Friday afternoon, that cited national security and barred the models from being accessed by any foreign national.
It is a striking sequence: a celebrated, benchmark-topping launch one moment, and a worldwide shutoff the next, triggered not by a technical failure but by a government order. The episode is one of the clearest signs yet that frontier AI models are now being treated by Washington much like sensitive defense technology — and it raises hard questions about how, and how transparently, that power will be used.
Editor's note: This is a developing story based on Anthropic's public statement and reporting from multiple outlets as of June 15, 2026. Some details below are attributed to single sources and may evolve. We will update as the situation develops.
What happened, in order
According to Anthropic's own account, the directive arrived on Friday, June 12, at 5:21 PM Eastern Time. It instructed the company to suspend "all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States" — language that, the company noted, extended even to its own foreign-national employees.
Crucially, Anthropic did not have a way to instantly verify the nationality of every user and block only foreign nationals at a moment's notice. To be certain it complied with a legal order, it took the blunt-instrument path: it disabled both models for everyone, everywhere. Hundreds of millions of users lost access to the two systems at short notice — not because they were foreign nationals, but because that was the only way to guarantee no foreign national could get through.
The company was careful to limit the blast radius. "Access to all other Anthropic models will not be affected," it said — including its widely used Claude Opus 4.8. Only Fable 5 and Mythos 5 went dark.
Why the government acted
The directive cited national security but, by Anthropic's description, did not spell out the specific concern in detail. Reading between the lines of the company's statement and outside reporting, the trigger appears to involve the models' cybersecurity capabilities.
Mythos 5 is understood to carry powerful abilities to detect software vulnerabilities — exactly the kind of dual-use capability that security officials worry could be turned into a cyberweapon "in the wrong hands," as Al Jazeera framed the long-standing concern. Fable 5, in this telling, functions as a more constrained, safety-guarded interface designed to keep users from directly tapping those deeper capabilities.
The specific spark, according to Anthropic, was that the government believed it had found a way to bypass — or "jailbreak" — Fable 5's safeguards. That is where Anthropic and Washington sharply diverge.
Anthropic's pushback
Anthropic is not going quietly. The company says it reviewed the government's demonstration and concluded the alleged jailbreak was far narrower than the dramatic response implied. It described the issue as involving "a small number of previously known, minor vulnerabilities," and characterized the technique as non-universal — essentially amounting to asking the model to read code and point out flaws, something many tools can do.
"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 wrote. It argued that its layered, "defense-in-depth" safety approach left the real-world risk comparable to other models already on the market, and it criticized the process itself as lacking transparency and technical grounding. Anthropic said it is working to restore access "as soon as possible."
In short: the company is framing this less as a genuine security emergency and more as an overreaction — or worse — to a routine, well-understood class of model behavior.
The business stakes are enormous
The timing could hardly be more sensitive. According to Fortune, Anthropic confidentially filed for a public listing earlier this month, reportedly at a valuation in the region of $965 billion. A surprise government action that yanks your flagship products offline — and signals regulatory unpredictability to would-be investors — is precisely the kind of headline a company does not want on the eve of an IPO. Fortune reported that analysts believe the directive "could make investors less enthusiastic about an Anthropic IPO."
There is a strategic dimension, too. Frontier models are the product. An order that can disable them overnight, applied through export-control authority, introduces a category of risk that AI investors have not had to price before: not market competition or a failed benchmark, but a letter from the government.
The bigger picture: this didn't come from nowhere
Reporting suggests the move fits into a tense, months-long relationship between Anthropic and the current administration. Per Fortune, the episode follows a February directive ordering federal agencies to stop using Anthropic's models after the company declined Pentagon contract terms that would have allowed deployment "for any lawful purpose," and a subsequent move in March designating Anthropic a "supply chain risk." (These elements come largely from single-source reporting and should be treated as context rather than settled fact.)
Whatever the backstory, the mechanism on display is the headline. Treating access "by any foreign national" as something requiring an export-control order is significant: under US law, making controlled technology available to a foreign national — even one standing inside the United States — can legally count as an "export." Applying that logic to a commercial chatbot used by the global public is a notable escalation in how AI is regulated, and a potential template for future actions against any frontier developer.
What the experts are saying
Outside voices, quoted by Fortune, captured the unease and the ambiguity:
- Dean Ball, an AI policy commentator, said he couldn't tell "if this is lawfare against Anthropic in particular or extreme national-security hawkery" — neatly summarizing the two competing interpretations.
- Peter Girnus, a cybersecurity researcher, offered a pointed warning about the industry's own rhetoric: "If you describe your product as a munition in every press release, eventually a government takes you at your word."
- Gary Marcus, a frequent AI-industry critic, cautioned that a directive blocking foreign nationals could push "Chinese-born AI researchers" to leave the US, undercutting the very competitiveness such controls are meant to protect.
Those three reactions map the debate well: Is this a legitimate security safeguard, a heavy-handed political maneuver, or a self-inflicted wound from an industry that markets its models as borderline weapons?
Why this matters beyond Anthropic
Strip away the specifics and three larger shifts are visible:
- Frontier AI is now treated like strategic technology. Export-control machinery built for chips, encryption, and weapons systems is being pointed at large language models. Every major developer should assume it could be next.
- "Foreign national" access is becoming a fault line. If providing model access to non-citizens can be deemed an export, AI companies face thorny questions about global products, international teams, and cloud access that have no clean technical answer.
- Process and transparency are now part of the story. Anthropic's loudest complaint isn't only the outcome but the manner — a late-Friday letter, thin on technical detail, with sweeping immediate effect. How governments wield this power, and how much they explain, will shape industry trust and investment.
What to watch next
The immediate question is whether Anthropic can get the order reversed or narrowed — and how fast. The company is openly disputing the technical basis, which sets up a test of whether a frontier developer can successfully challenge a national-security designation. Watch also for any official government statement (notably absent so far), for whether other developers' models draw similar scrutiny, and for how investors weighing Anthropic's reported IPO read the episode.
The bottom line
In the space of a few days, Anthropic went from celebrating a landmark launch to switching it off worldwide on government orders — over what the company insists is a narrow, well-known class of vulnerability. Both things can be true at once: governments have a legitimate interest in frontier AI's dual-use risks, and a sweeping, lightly-explained order that disables a product used by hundreds of millions sets an uneasy precedent.
Either way, a line has been crossed. The question of who gets to use the most powerful AI models — and who gets to decide — is no longer abstract. It now arrives by letter, on a Friday afternoon, with national security on the letterhead.
Reporting & sources
This article draws on Anthropic's official statement and reporting from multiple outlets. Single-sourced claims (valuation, prior-policy backstory, and expert quotes) are attributed inline and should be treated as developing.
- Statement on the US government directive to suspend access to Fable 5 and Mythos 5 — Anthropic
- Anthropic disables Fable and Mythos AI models following U.S. government export ban — Fortune
- US orders Anthropic to disable AI models for all foreign nationals — Al Jazeera
- Anthropic Pulls Its Most Powerful AI Models After U.S. Bars Foreign Access — TIME
- Anthropic disables access to Fable 5 and Mythos 5 to comply with government directive — CNBC
