Anthropic Disables Fable 5 and Mythos 5 — Three Days After Launch

Anthropic disabled Claude Fable 5 and Claude Mythos 5 on June 12, 2026, three days after launching them on June 9. The shutdown is global and applies to every customer — the Claude apps, the API, and third-party surfaces including GitHub Copilot. This is not a deprecation on the normal schedule; it is an abrupt, directive-driven suspension that Anthropic disclosed in its own statement and that CNBC, 9to5Mac, and other outlets corroborated. Operators who moved workloads onto claude-fable-5 in the last three days need to act now.

Why it happened: a US export-control order. In its own statement, Anthropic said the US government, "citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." Because it cannot reliably separate foreign nationals from the rest of its user base in real time, Anthropic disabled both models for everyone — its stated position is that "the net effect of this order is that Anthropic must abruptly disable Fable 5 and Mythos 5 for all customers to ensure compliance." Anthropic added that the government's letter gave no specific details of the concern, but said its understanding is that the issue is a demonstrated jailbreak of Fable 5 — a technique it characterized as "narrow" and "non-universal" and disputed as grounds for a recall. It stated: "We are complying with the government's legal directive and are removing access to Fable 5 and Mythos 5 for all users. However, 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."

What still works — and what falls back. Anthropic confirmed that all other models are unaffected: Opus 4.8, Sonnet, and Haiku continue to run normally. For workloads that were pointed at Fable 5, the behavior is now two-fold. Existing Fable 5 sessions terminate with an error. New requests route to whichever model your integration specifies as a default, or to Claude Opus 4.8 in the Claude apps. This is the same Opus 4.8 that Fable 5's built-in safeguards already fell back to for sensitive cybersecurity, biology, chemistry, and distillation queries — so most teams already have a working fallback path, whether or not they planned for one.

The billing picture flips, at least for now. At launch, the headline operator action item was the June 22 cutoff, after which Fable 5 would require usage credits beyond subscription coverage. That deadline is moot while the model is suspended. There is a small silver lining on cost: Opus 4.8 is priced at $5/$25 per million input/output tokens — half of Fable 5's $10/$50 — so any workload that reverts to Opus 4.8 gets cheaper, not more expensive. The capability ceiling drops back to Opus 4.8, but the bill does too.

GitHub Copilot pulled it the same day. Fable 5 had been generally available in GitHub Copilot as of June 9. Effective June 12, access to Fable 5 was suspended across all Copilot experiences in line with Anthropic's announcement. Any Copilot configuration that had been switched to Fable 5 reverts to a different available model — teams should confirm which one is now serving their completions and chat.

Is this permanent? As of June 12, neither Fable 5 nor Mythos 5 has been placed on Anthropic's formal model-retirement schedule — meaning Anthropic has not committed to killing them permanently. Anthropic says it is "working to restore access as soon as possible" and frames the episode as likely "a misunderstanding," but it has given no timeline, and both models are unavailable everywhere right now. Treat the suspension as indefinite for planning purposes: do not build the next sprint around Fable 5 being back by a specific day.

The operator lesson. The three-day round trip from frontier launch to global recall is a vivid reminder that a brand-new model — however capable — is not a stable dependency. Regulatory, safety, and policy actions can remove a model from production with no notice and no grace period. The teams least disrupted today are the ones that pinned a known-good default (Opus 4.8 or Sonnet) and treated Fable 5 as an opt-in upgrade rather than a hard dependency. The teams scrambling are the ones that rewired production model routing to a three-day-old model.