How Do You Find an Illegal Image Without Looking at It?
Summary
The article explains how large-scale CSAM detection uses perceptual hashing (PDQ/PhotoDNA) to fingerprint known material, while unknown or AI-generated content is tackled with classifiers in a layered Hasher-Matcher-Actioner approach. It covers TMK for video, the distinction between known and unknown CSAM, privacy considerations, and practical tooling used by platforms to scale detection without exposing original images.