CopyJudge: Automated Copyright Infringement Identification and Mitigation in Text-to-Image Diffusion Models
This addresses copyright disputes for users and developers of text-to-image diffusion models, offering an incremental improvement in automated infringement detection and mitigation.
The paper tackles the problem of identifying copyright infringement in AI-generated images by proposing CopyJudge, an automated framework that uses large vision-language models to assess substantial similarity between copyrighted and generated images, achieving comparable state-of-the-art performance. It also introduces a mitigation strategy that optimizes prompts to avoid infringement while preserving content, effectively reducing memorization and IP issues with high alignment.
Assessing whether AI-generated images are substantially similar to source works is a crucial step in resolving copyright disputes. In this paper, we propose CopyJudge, a novel automated infringement identification framework that leverages large vision-language models (LVLMs) to simulate practical court processes for determining substantial similarity between copyrighted images and those generated by text-to-image diffusion models. Specifically, we employ an abstraction-filtration-comparison test framework based on the multi-LVLM debate to assess the likelihood of infringement and provide detailed judgment rationales. Based on these judgments, we further introduce a general LVLM-based mitigation strategy that automatically optimizes infringing prompts by avoiding sensitive expressions while preserving the non-infringing content. Furthermore, assuming the input noise is controllable, our approach can be enhanced by iteratively exploring non-infringing noise vectors within the diffusion latent space, even without modifying the original prompts. Experimental results show that our automated identification method achieves comparable state-of-the-art performance, while offering superior generalization and interpretability across various forms of infringement, and that our mitigation method more effectively mitigates memorization and IP infringement with a high degree of alignment to the original non-infringing expressions.