The White House's emergency export controls on Anthropic's flagship AI model, Mythos, were triggered by suspicions that a China-linked group had already obtained access to it, according to a Semafor exclusive.
Since its release in April, Mythos has been under extremely tight restrictions, available only to a handful of enterprise partners, as Anthropic tried to patch security holes and keep it out of hackers' hands. The model is considered a "cyber-weapon-grade" threat by US officials because it can autonomously discover and exploit code vulnerabilities. Washington fears that if the Chinese government gets Mythos, it could reverse-engineer and replicate it using techniques like "distillation" — posing a national security risk. Anthropic responded by saying the White House never mentioned China getting access in previous discussions, and stressed that the company blocks all product access from within China.