Fable 5 Shutdown: US Government Order Pulls Anthropic’s Most Powerful Claude Model

by Zaira Varghese 15. June 2026 AI 0

The Fable 5 shutdown became the most talked-about AI story of the week after Anthropic abruptly disabled all public access to its two most powerful models, Claude Fable 5 and Mythos 5, to comply with a US government export control directive citing national security. Both models had launched only three days earlier, on June 9, making the takedown one of the most dramatic regulatory interventions the AI industry has seen.

What triggered the Fable 5 shutdown

According to CNBC’s reporting, the Commerce Department issued the directive at 5:21pm ET on June 12, ordering Anthropic to suspend access to Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States, including the company’s own foreign-national employees. Reporting from Axios and CNBC indicates the administration moved after another company claimed it had found a way to jailbreak Mythos, raising alarm inside the government about how exploitable the models might be in the wrong hands.

Anthropic framed the situation as a misunderstanding and said it was working to restore access as soon as possible. But the practical effect was immediate and total: the two flagship models simply went dark worldwide.

Why Anthropic pulled the models for everyone

The directive technically targeted foreign nationals, not US users. The problem is that Anthropic cannot reliably filter foreign nationals from American users in real time across its global infrastructure. Faced with a compliance order it could not partially satisfy, the company chose the only option that guaranteed compliance: shut both models down for everyone. Crucially, every other Anthropic model stayed online. Claude Opus 4.8 and the rest of the lineup were unaffected, so most everyday users could keep working while the highest-capability tier vanished.

The jailbreak and system-prompt leak behind the panic

The shutdown did not happen in a vacuum. A day after launch, a well-known red-teamer operating as “Pliny the Liberator” claimed to have bypassed Fable 5’s safety classifiers using a coordinated multi-agent technique, then published roughly 120,000 characters of the model’s internal system prompt on GitHub. Security researchers later argued the demonstration was narrower than the viral headlines suggested, but the spectacle of a freshly launched frontier model being publicly probed fed directly into the national-security anxiety that produced the order.

Fable 5 launched June 9 and was pulled June 12.
The directive cited export-control and national-security authorities.
Mythos 5 was suspended alongside Fable 5; Opus 4.8 stayed live.
Anthropic is processing refunds for affected paid customers.

What the Fable 5 shutdown means for enterprises

For businesses that had already wired Fable 5 into production workflows, the Fable 5 shutdown was a hard lesson in vendor risk. Sessions routing to the model began failing, with queries falling back to older, less capable systems and no automatic path that preserved output quality. The episode reignited a debate developers are calling “hardware sovereignty” — the idea that the most critical AI workloads may need self-hosted or open-weight fallbacks that no single government order can switch off. Multi-provider routing, once a nice-to-have, suddenly looked like operational necessity.

A standoff that is far from over

The order is the latest escalation in a months-long conflict between Anthropic and Washington that already includes an active lawsuit over a separate “supply chain risk” designation. No firm restoration date for Fable 5 has been given, and the longer the models stay offline, the more the Fable 5 shutdown looks less like a one-off compliance hiccup and more like a preview of how governments may police frontier AI going forward. For an industry that has spent two years assuming its most powerful tools would always be a click away, that is a sobering shift.