IP Protection in the Era of Visual Generative AI: A Survey

A survey on IP protection for visual generative AI proposes a two-dimensional taxonomy: the primary dimension divides by the risk variable being regulated into information exposure control, generation behavior constraint, and attribution and accountability, and the secondary dimension distinguishes data IP from model IP. This framework systematically organizes various protection methods, aligns evaluation protocols with protection goals, and discusses open challenges such as model-level proactive protection, evaluation standardization, robustness against adaptive attacks, and interpretable evidence.