Amazon Researchers Behind Fable 5 Jailbreak? Report Links Amazon to Anthropic AI Export Control Crisis

The U.S. government’s sudden decision to block foreign access to Anthropic’s most advanced AI models — Fable 5 and Mythos 5 — has taken a new twist. According to reports, the catalyst was research conducted by Amazon researchers who successfully demonstrated a jailbreak on the models.

This development adds another layer of complexity to the already tense relationship between major AI labs and increasing government scrutiny over frontier AI systems.

What Happened

Late last week, the U.S. Department of Commerce issued an export control directive ordering Anthropic to restrict access to Fable 5 and Mythos 5 for any foreign nationals — including Anthropic’s own non-U.S. employees.

To comply, Anthropic was forced to disable access to both models for all users worldwide, not just foreign ones. The company stated that the government acted after becoming aware of a method to “jailbreak” Fable 5.

New reporting now indicates that Amazon researchers were the ones who demonstrated this jailbreak technique to the government.

The Jailbreak Technique

According to sources familiar with the matter, Amazon researchers used a series of carefully crafted prompts to bypass Anthropic’s safeguards. The goal was to make the model reveal information about security vulnerabilities — essentially using it for code auditing and vulnerability discovery.

Anthropic has pushed back on the severity of the findings, stating that:

• The vulnerabilities were minor and previously known.
• Similar issues exist in other publicly available models.
• The jailbreak was narrow rather than a universal bypass.

Despite Anthropic’s position, the U.S. government viewed the demonstrated capability as a national security concern — particularly because Mythos-class models were designed with powerful cybersecurity and reasoning abilities.

Why This Matters

This incident highlights several important developments in the AI industry:

1. Intensifying Competition Between AI Labs The fact that Amazon researchers reportedly played a role in triggering regulatory action against Anthropic underscores the high-stakes competition between major AI companies.
2. AI Safety as a Geopolitical Issue What began as a technical safety concern (jailbreaking) quickly escalated into an export control decision. This shows how AI capabilities are increasingly being treated as strategic national assets.
3. Growing Government Involvement The U.S. is taking a more aggressive stance on controlling access to the most powerful AI models, even within domestic companies.
4. Challenges of “Safe” Model Releases Anthropic had positioned Fable 5 as a more restricted, safety-focused version of its powerful Mythos model. The fact that it could still be jailbroken raises questions about how effectively companies can contain dangerous capabilities.

Anthropic’s Response

In its official statement, Anthropic maintained that the vulnerabilities were limited in scope and that other leading models (including those from competitors) have similar weaknesses. The company expressed disagreement with the government’s decision while confirming it would comply with the legal directive.

Anthropic also noted that it is working to restore access as quickly as possible and is engaging with officials to resolve the issue.

Broader Implications

This episode is part of a larger trend in 2026 where:

• Governments are treating advanced AI models as dual-use technologies requiring export controls.
• Competition between AI companies is spilling into regulatory and national security domains.
• The line between AI safety research and competitive intelligence is becoming increasingly blurred.

It also raises questions about how future “jailbreak” discoveries will be handled — whether they will be disclosed responsibly or used as leverage between competing organizations and governments.

What’s Next

Anthropic has indicated it expects to provide more updates within 24 hours. Industry observers will be watching closely to see:

• Whether access to Fable 5 and Mythos 5 is restored quickly.
• How other AI labs respond to increased government scrutiny.
• Whether this leads to more formal regulations around AI model safety and access.