After 31M Tests, 'Adversarial' Patterns Successfully Evade AI Surveillance Cameras

alexbilz · x · 2026-08-10

Bill Swearingen spent a year and conducted around 31 million tests to develop noRecognition, an adversarial pattern project.

How it works: Applying these computer-generated patterns to clothing or objects can effectively prevent AI detection algorithms in commonly deployed license plate readers and surveillance cameras across the U.S. from recognizing the covered objects.
Impact: The technology allows individuals to evade automatic detection and algorithmic surveillance, sparking debates over privacy protection versus public safety.

More from Safety