The US government imposed export controls on Anthropic's advanced AI models 'Mythos' and 'Fable' citing national security risks. Anthropic had 90 minutes to roll back access and is now negotiating with the White House.
The US government imposed export controls on Anthropic's advanced AI models 'Mythos' and 'Fable', citing national security risks. The Commerce Department gave Anthropic 90 minutes to block access to both models, which the company complied with.
Fable was a 'safe' version of Mythos with guardrails against biological weapons and cybersecurity threats, but Amazon researchers jailbroke it within days. Amazon CEO Andy Jassy reported this to the Commerce Department, triggering the export controls. Semafor reported that China may have accessed Mythos. Anthropic CEO Dario Amodei had long marketed Mythos as 'too dangerous to release', but its security proved flimsy.
This incident highlights the clash between AI safety rhetoric and export control policy. Anthropic is now negotiating with the White House, but the government's强硬 stance suggests low trust in self-regulation. It also shows how marketing AI as 'too powerful' can backfire, inviting stricter oversight.