Your AI’s Secret Thoughts Were Never Actually Secret #ai #programming

A short explainer on a security researcher trick that reportedly extracts encrypted chain-of-thought reasoning by replaying it through a smaller, easier-to-jailbreak sibling model; for developers int…