Tech to protect images against AI scrapers can be beaten

Computer scientists have created LightShed, a method to counteract image protection techniques like Glaze and Nightshade used by artists to prevent AI training on their work. LightShed detects and removes data poisoning patterns, potentially allowing AI models to access copyrighted images. The researchers aim to highlight weaknesses in existing protection schemes, raising concerns about current defenses against non-consensual AI training and possibly leading to improved protection techniques.