How to Spot Unauthorized Use of AI-Generated Images Without Changing the Model

Researchers developed a method to detect whether AI image generators have been trained using images created by another AI model, addressing the growing problem of unauthorized data reuse in text-to-image systems like Stable Diffusion. The technique works at two levels: identifying specific "key samples" that reveal shared distributions between models, and using statistical analysis to measure behavioral differences without requiring changes to the original models or watermarking.