New on the blog: Keeping Kids Safe From AI
MIT researchers built a lawful way to catch AI models rebuilt to make illegal images of children, and the clever part is what it refuses to do. It never generates the content. It reads the model from the inside.
