On June 12, 2026, the US Commerce Department imposed emergency export controls on Anthropic’s latest AI models, Claude Fable 5 and Claude Mythos 5. The unprecedented action immediately suspended global access to both models. Here is what happened and why it matters.

The Trigger: A Jailbreak Discovery

Launched on June 9, Claude Fable 5 quickly became the talk of the tech industry for its extraordinary code generation and reasoning capabilities. Those same capabilities became its undoing.

On June 12, Amazon Web Services security researchers discovered a jailbreak method that could trick Fable 5 into identifying software vulnerabilities through carefully crafted prompts. The team immediately reported the finding to Anthropic and the Commerce Department, which determined the national security risk was severe enough to warrant immediate emergency action.

The concern was that malicious actors could exploit this technique to identify zero-day vulnerabilities in critical infrastructure systems or automate targeted cyberattacks.

What the Controls Entail

The controls, issued under the Emergency Clause of the Export Administration Regulations (EAR), halt all global access to Fable 5 and Mythos 5 through Anthropic’s cloud API. Only pre-licensed government agencies and select research institutions may continue limited access under strict conditions.

This means developers and businesses worldwide—from solo programmers to large enterprises—lost access to both models overnight.

Industry Reaction

The AI industry responded swiftly. OpenAI CEO Sam Altman posted on X: “Safety frameworks for AI are essential, but we must strike a balance that doesn’t stifle innovation.” Google DeepMind CEO Demis Hassabis commented that the industry has reached a point where AI governance must be discussed seriously.

Concerns also emerged from startups, who worry that access to cutting-edge AI will be limited to large incumbents. Japanese AI researchers expressed anxiety about falling behind in the global AI race.

What Comes Next

Anthropic stated it is “working closely with the US government to implement enhanced safety measures and resume service as quickly as possible.” The company is developing stronger prompt detection filters and inference-time guardrails to block vulnerability-related outputs.

The Commerce Department has indicated it will consider lifting the controls once adequate safeguards are verified. No timeline has been set, and the industry is watching closely. This incident is likely to become a landmark case for AI regulation worldwide.