The Core Shift: From Open Access to Export Controls

On June 9, 2026, Anthropic released a modified AI model called Fable, intended to be safer than its predecessor Mythos. Within days, the US government imposed export controls, citing national security threats. Anthropic revoked access to both models. This sequence marks a pivotal moment in AI governance: the first time a frontier AI model has been treated like a controlled munition.

The trigger? Amazon CEO Andy Jassy personally informed government officials that Fable posed a cybersecurity danger. Amazon is both an investor in Anthropic and a competitor building its own AI models. This conflict of interest raises questions about the integrity of the regulatory process.