• How the crisis began
• Shutdown and immediate fallout
• Government and industry perspectives
• Diplomatic, geopolitical, and commercial shockwaves
• Escalation to crisis talks
• Broader debate: safety vs. control

US Government Orders Anthropic to Shut Down Fable 5 and Mythos 5 Models The world’s most capable publicly available coding model went dark just three days after launch, as a clash between Anthropic and the U.S. government over security, sovereignty, and control of frontier AI models spilled into public view.

§How the crisis began

Anthropic released Fable 5 and its higher-powered sibling Mythos 5 on June 9, describing Fable as its most capable generally available model and positioning Mythos for tightly controlled cybersecurity use. Within days, Fable 5 was topping benchmarks and “dominated every benchmark” against rivals like GPT 5.5.

Behind the scenes, however, U.S. officials were growing uneasy. Anthropic CEO Dario Amodei had previously warned that Mythos posed serious national security risks, and federal agencies were already experimenting with it for cyber defense.

On June 12, after Amazon security researchers reported they could prompt Fable 5 to surface code vulnerabilities, Amazon CEO Andy Jassy relayed the findings to senior officials. Commerce Secretary Howard Lutnick then sent Anthropic an export-control letter subjecting Fable 5 and Mythos 5 to strict controls for all foreign nationals, including Anthropic’s own staff. Anthropic says the government provided only “verbal evidence of a potential narrow, non-universal jailbreak.”

§Shutdown and immediate fallout

With no practical way to separate U.S. and foreign users, Anthropic “abruptly disable[d] Fable 5 and Mythos 5 for all our customers” while disputing that such a narrow exploit justified a recall. The company warned that if this standard were applied broadly, it “would essentially halt all new model deployments for all frontier model providers.”

The shutdown instantly reshaped the market: Fable 5, briefly the top model on public leaderboards, vanished, leaving GPT 5.5 as “the strongest model available” largely by default.

§Government and industry perspectives

Administration officials argued they were treating cutting-edge AI as a national security asset and said the models must remain locked down until the “national security apparatus is hardened” against such threats. Reports also described fears that a China-linked group might have accessed Mythos, raising the risk of reverse engineering through distillation.

Anthropic countered that Fable’s safeguards had been extensively red-teamed and were “substantially more effective than those of any previously deployed model,” and that no one had demonstrated a universal jailbreak.

Cybersecurity leaders quickly pushed back against the ban. In an open letter, roughly 100 experts argued that “pulling the best AI tools from defenders while adversaries keep building is not safety, it is sabotage,” warning the move “has taken the best models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it.” Axios reported that experts like former Facebook CSO Alex Stamos stressed that the same kind of exploit exists “across other leading AI models, too.”

§Diplomatic, geopolitical, and commercial shockwaves

Abroad, the sudden cutoff fueled a broader debate over AI sovereignty. In India, Anthropic’s second-largest market, access to its most advanced models disappeared overnight, just a day after a major partnership with Tata Consultancy Services, triggering calls for a multi-billion-dollar sovereign AI fund and greater reliance on open-source models. Commentators argued the shutdown “handed India’s sovereign AI movement its strongest argument yet.”

In Europe, French startup Mistral framed the incident as proof that relying on U.S. providers leaves others vulnerable to Washington’s politics, reinforcing its pitch for open-weight, locally hosted European models. Analysts said the restrictions were “a win for Europe’s top AI startup” by validating its sovereignty message.

Canadian Prime Minister Mark Carney likened the episode to the 2008 financial crisis, calling it a warning about “over-reliance on certain models” and urging redundancy and diversity in AI infrastructure.

Meanwhile, critics warned that cutting off Anthropic’s classified Mythos system from many foreign partners was “a gift to China,” strengthening Beijing’s pitch for its own competing AI stack. Semafor reporting, summarized by The Verge, said U.S. fears over potential Chinese access were part of the calculus behind export restrictions on Mythos.

§Escalation to crisis talks

Over the weekend and into June 15, the dispute evolved into a full-blown confrontation between Silicon Valley and the Trump administration. Axios reported that Anthropic now sat in the paradoxical position of being on a Pentagon blacklist as too risky for U.S. use while also deemed too dangerous for foreign access.

Behind closed doors, senior officials and Anthropic executives held hours-long calls, followed by emergency trips to Washington by the company’s top technical staff. Business Insider described “tense” conversations as both sides sought a truce, with Anthropic presenting its safeguards and red-teaming methodology in detail. The Commerce Department’s Center for AI Standards and Innovation led technical sessions aimed at finding conditions for restoring access.

§Broader debate: safety vs. control

For many in the AI and security communities, the Fable 5 episode crystallizes a deeper tension: how to police dangerous capabilities without letting national governments unilaterally dictate who may use frontier AI.

TechCrunch argued the enforcement letter “should be a wake-up call” that the U.S. government can now “shut you and your products down” without court approval, and suggested personality clashes and prior disputes over surveillance and lethal uses of AI may have shaped the response as much as the technical issue itself.

At the same time, Anthropic itself has long warned that models like Mythos can pose grave risks if misused, and previously kept it restricted via programs like Project Glasswing for vetted defenders. Supporters of the crackdown say this proves why governments must err on the side of caution.

As crisis talks continue in Washington, Fable 5 and Mythos 5 remain offline. Developers and defenders worldwide are left to work with less capable systems, while governments and companies race to redefine who ultimately holds the power to flip the off switch on the most advanced AI.

Continue reading https://foxvector.com/stories/019ecedc-560e-0293-718e-333924e81bdd