The US Government Just Pulled Claude Fable 5. Here's What Actually Happened.

The fastest product launch and shutdown in AI history happened last week, and if you missed it, here is the full story.

On June 9, 2026, Anthropic released Claude Fable 5 — its first publicly available Mythos-class model, a tier that sits above even the Opus class. It was the most capable model Anthropic had ever shipped to the public, ranking number two on the BenchLM.ai provisional leaderboard with an 80.3% score on SWE-Bench Pro. Three days later, on June 12 at 5:21 PM Eastern, the US government issued an export control directive and Anthropic pulled both Fable 5 and Mythos 5 for every customer on the planet.

Not just for foreign users. For everyone.

Key takeaways

• Claude Fable 5 launched June 9, 2026 and was pulled three days later by a US government export control directive citing national security
• The government believed a jailbreak method had been discovered that could allow Fable 5 to generate restricted outputs
• Anthropic could not filter foreign nationals from US users in real time, so it shut down both models globally to comply
• Claude Opus 4.8 remains available and is serving as the fallback on all Anthropic plans
• No public timeline has been given for Fable 5 or Mythos 5 to return

What was Claude Fable 5, exactly?

Fable 5 was the first model from Anthropic's Mythos class to reach general availability. Think of Mythos as the tier above Opus — the company's internal frontier that it had never before made available to API users or consumer subscribers.

The numbers were striking. On SWE-Bench Pro, Fable 5 hit 80.3%, while Claude Opus 4.8 sits at 69.2% and GPT-5.5 at 58.6%. On the Artificial Analysis Intelligence Index, Fable 5 reached a score of 64.9 — roughly five points ahead of GPT-5.5. For coding tasks specifically, it led all eight major coding benchmarks by an average of 11.8 points while costing 72% less per output token than GPT-5.5 Pro.

Pricing was set at $10 per million input tokens and $50 per million output tokens. That is expensive compared to Claude Sonnet 4.6 ($3/$15), but meaningfully cheaper than OpenAI's GPT-5.5 Pro variant at $30/$180.

Why did the government pull it?

The directive came from US national security authorities and named both Fable 5 and Mythos 5. The stated concern, based on Anthropic's public statement posted to their site on June 12, is that the government believed it had become aware of a jailbreaking method — a way to bypass the model's safety filters that could allow it to generate outputs that are restricted under export control law.

Anthropically could not verify or deny the specific vulnerability before complying, and by law, the directive required immediate action.

The export control angle is important. This is not a recall for a bug in the traditional sense. It is closer to the government stepping in the way it might step in on advanced semiconductor exports. The concern is about what a sufficiently capable model could help foreign actors do — particularly in domains like bioweapons design, cyberattack planning, or advanced materials synthesis.

Why did Anthropic shut it down for everyone, not just foreign users?

This is the most operationally interesting detail of the whole situation. Anthropic posted on their official channels that the order required suspending access for any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The problem is that Anthropic, like every major AI company, does not verify nationality at the account level in real time. It cannot separate a US citizen subscriber from a Canadian one, or a Brazilian researcher, or a Korean developer, on the fly. So to achieve compliance with a directive that had no grace period, the company turned off both models for the entire world simultaneously.

This is a reminder that AI infrastructure is increasingly subject to the same kinds of regulatory frameworks as physical technology exports. The AI model you're building on today can be taken away not because it stopped working, but because of a government directive you never saw coming.

What can you use instead right now?

If you were using Claude Fable 5 or Mythos 5 via the API or Claude.ai subscriptions, your account has automatically fallen back to Claude Opus 4.8. Anthropic has made this clear in its communication.

Claude Opus 4.8 is not a downgrade in any painful sense — it still scored 61.4 on the Artificial Analysis Intelligence Index, which made it the first model to break 60 on that index when it launched in May 2026. It leads several coding benchmarks and is considered the best available model from Anthropic as of today.

If you need a strong frontier alternative while the situation with Fable 5 is unresolved:

• GPT-5.5 from OpenAI is available at $5/$30 per million tokens and scores 89 on BenchLM's overall rankings
• Gemini 3.5 Flash is the most affordable frontier-class option at $1.50/$9 per million tokens, with particular strength on agentic and multimodal tasks
• DeepSeek V4 Pro remains available as an open-weight option at dramatically lower API costs

What happens next?

Anthropically has said it is working with the government to understand the specific concern and determine a path forward. The company's statement does not include a timeline, and as of June 18, both models remain offline.

This is genuinely new territory for the AI industry. No model of this capability class has ever been suspended by government order this quickly after launch. It raises questions that will matter for months or years ahead: How do AI companies comply with export control frameworks designed for physical hardware when software models can be copied and distributed? How do you perform nationality checks at model-inference time? And what does it mean for enterprise customers building products on top of these APIs when that foundation can vanish in an afternoon?

For now, the most practical thing you can do if you depend on Anthropic is to test your workflows against Claude Opus 4.8 and make sure your system prompts are not Fable-5-specific. The capabilities are real but the access is not guaranteed.

Frequently asked questions

Is Claude Fable 5 still available anywhere?

No. As of June 18, 2026, Claude Fable 5 and Claude Mythos 5 are suspended globally. Anthropic took them offline on June 12 following a US government export control directive and has not announced a return date. All accounts have been automatically moved to Claude Opus 4.8.

What triggered the US government to pull Claude Fable 5?

The government cited national security authorities and indicated it had become aware of a method to jailbreak Fable 5 in ways that could enable restricted outputs. The specific jailbreak has not been publicly disclosed. The directive was not tied to any specific incident that Anthropic has confirmed publicly.

What is the difference between Claude Fable 5 and Claude Opus 4.8?

Fable 5 was Anthropic's Mythos-class model, a tier above Opus — its highest-capability model ever released publicly. On SWE-Bench Pro, Fable 5 scored 80.3% versus Opus 4.8's 69.2%. Fable 5 was priced at $10/$50 per million tokens versus Opus 4.8's standard pricing. Opus 4.8 is still the recommended frontier option from Anthropic while Fable 5 remains suspended.