The US Government issued an export control directive prohibiting foreign nationals from accessing Claude Fable and Claude Mythos. These latest models are now restricted due to cybersecurity risks. This move follows a third-party assessment conducted by Amazon, a trusted government contractor, which identified specific security vulnerabilities that justify these strict access limitations.

CEO Dario Amodei previously argued that the government should block model deployment if third-party assessments show unacceptable risks. He specifically cited cybersecurity and biological weapons as primary concerns. This directive effectively grants the government the exact power Amodei requested in his official policy on AI exponential growth, turning his own regulatory arguments into a legal reality.

Critics argue Anthropic likely expected these regulations to target open source projects or smaller competitors rather than themselves. By framing their models as national security concerns, the company created a legal record that makes challenging these restrictions difficult. The current restrictions reflect a literal application of the company's own public stances on AI safety.

This situation demonstrates the risk of inviting government oversight without defining strict boundaries. Anthropic now operates under the same restrictive framework it advocated for years. The government is simply treating the company's warnings about biological and cyber risks as factual evidence for its export controls.