US cyber agency is using Anthropic’s Mythos to audit government code, sources say

A federal defender is reportedly pointing a private, offensive-grade AI model at the government’s own software, though almost nothing about the work is on the record.