Claude Fable 5 Is Offline -- US Export Controls and the Jailbreak Behind the Ban

The Claude Fable 5 shutdown was abrupt: Anthropic's most capable publicly available AI model launched on June 9, 2026, and was gone from every user's account by June 12. On that evening, the US Department of Commerce issued an export control directive ordering Anthropic to suspend all access to Fable 5 and Claude Mythos 5 for any foreign national, whether inside or outside the United States. Unable to verify user nationality across hundreds of millions of accounts in real time, Anthropic disabled both models globally. As of June 16, neither is back.

What Is an Export Control Directive -- and Why Did One Hit an AI Model?

An export control directive is a legally binding order from the US Department of Commerce that restricts the transfer of specific technologies to foreign nationals. These controls are standard in defense hardware, semiconductors, and encryption software. Applying one to a commercial AI model available to anyone with a credit card is unprecedented.

The legal authority sits with the Bureau of Industry and Security (BIS), which administers the Export Administration Regulations (EAR). Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei on June 1 flagging concerns, and the formal directive arrived June 12 at 5:21 pm ET, according to reporting by Nextgov/FCW.

What Amazon Found -- and Told the White House

The trigger was not a foreign intelligence agency or a public researcher. It was Amazon.

According to Fortune's June 14 reporting, Amazon CEO Andy Jassy raised concerns directly with senior Trump administration officials after Amazon researchers discovered that Fable 5 could surface security vulnerabilities in at least four software programs when fed a specific sequence of prompts. A separate "highly credible, trusted partner of both Anthropic and the government" independently identified a bypass in Fable 5's safety guardrails.

Trump AI adviser David Sacks said the administration asked Dario Amodei to fix the issue or withdraw the model before the directive was issued; Amodei reportedly refused, arguing the jailbreak was too narrow to warrant that response. The administration then issued the export control order.

Anthropic confirmed receiving the directive but disputed its severity. The company said no disclosed bypass had produced a harmful result or demonstrated a capability unique to Mythos compared with other publicly available models.

Why the Claude Fable 5 Shutdown Went Global

The directive's scope was foreign nationals -- not every user. So why did everyone lose access?

Because Anthropic cannot reliably distinguish US citizens from foreign nationals across hundreds of millions of accounts in real time. Partial compliance would mean routine violations of the directive. Rather than accept that risk, Anthropic issued a global shutoff.

In its public statement, Anthropic warned that applying this standard across the AI industry would "essentially halt all new model deployments for frontier model providers" -- a pointed argument against the directive's breadth. The company said it is seeking a deal with the administration to lift or narrow the restrictions.

The Precedent No AI Lab Wanted to Set

The Fable 5 incident is the first time a government has pulled a frontier commercial AI model from the market via export controls, and it exposes a structural vulnerability that every major AI lab shares: a model's deployment can be reversed overnight, with no grace period, at government request.

For API users, the impact was immediate. Fable 5 launched at $10 per million input tokens and $50 per million output tokens -- and developers who had begun integrating it lost access within three days, with no rollback path. Workflows built on a single AI provider have no buffer when that provider goes dark.

Multi-model setups face a simpler problem: switch to the next model and keep working. Tools like ByteChat that route queries across Claude, GPT, Gemini, and others via your own API keys let users redirect traffic without rebuilding a workflow. Broader than any one tool, the Fable 5 episode is a reminder that AI provider diversification is not just a cost question -- it is a resilience question.

The next government action against a frontier AI model will likely be faster and broader, because this one set the playbook.

Frequently asked questions

Is Claude Fable 5 still offline on June 16, 2026?

Yes. As of June 16, 2026, Claude Fable 5 and Claude Mythos 5 remain offline globally following the June 12 export control directive. Anthropic says it is negotiating with the US government but has not announced a deal or a restoration timeline. All other Claude models -- Opus 4.8, Sonnet 4.6, Haiku 4.5 -- remain fully available on the API and on Claude.ai.

What jailbreak caused the US government to shut down Claude Fable 5?

Amazon researchers found that Fable 5 could surface security vulnerabilities in at least four software programs when given a specific sequence of prompts. A separate third party also identified a bypass in the model's safety guardrails. Anthropic disputes the severity, calling the method narrow, non-universal, and comparable to what other public models can produce. The government has not released technical details of the jailbreak publicly.

Does the US export control directive affect the Claude API?

The directive is specific to Claude Fable 5 and Claude Mythos 5. The Claude API remains operational for Sonnet 4.6, Opus 4.8, Haiku 4.5, and all other models. Developers using non-Fable model IDs in their API calls are unaffected.