Algorithm Creates Real-World Patterns to Hide People from AI Cameras

Researchers unveiled an algorithm that generates physical, adversarial patterns—applied to clothing or objects—that fool AI surveillance cameras across multiple tasks (facial recognition, person and vehicle detection). It could bolster privacy but also enable evasion by criminals, and updating deployed cameras is harder than software patches, sparking an ongoing arms race between pattern designers and detection systems.