Anthropic's Mythos and Fable AI models trigger US government export controls

Anthropic, the AI company behind Claude, recently found itself in a conflict with the US government over its new AI models Mythos and Fable. Mythos, built in April, was so proficient at coding that Anthropic flagged it as a potential global cybersecurity threat, sharing it only with a small group of experts. In June, they released a modified, safer version called Fable to the public. The federal government quickly responded by placing export controls on Fable, deeming it a national security risk. Anthropic subsequently revoked access to both models. Amazon CEO Andy Jassy reportedly informed government officials about the danger, raising questions about Amazon's competing AI interests.

This incident has broader implications. European leaders, like France's Bruno Retailleau, see it as a wake-up call to build independent AI capabilities. However, the rise of capable and cheap Chinese open-source models (e.g., Zhipu) complicates this, as they offer fewer restrictions but also less safety. Security experts argue that restricting Anthropic's models could actually increase vulnerabilities by limiting defensive research. The episode also puts pressure on US lawmakers to create clearer AI regulations, as the current ad-hoc approach creates uncertainty for both companies and users.