CopyJudge: Automated Copyright Infringement Identification and Mitigation in Text-to-Image Diffusion Models

2025-02-21

AI Summary

This paper addresses copyright infringement risks in text-to-image diffusion models, where generated outputs may substantially resemble copyrighted images. We propose the first automated framework for judicial substantial similarity assessment and infringement mitigation. Methodologically: (1) we design a multi-visual-language-model (LVLM) debate mechanism that emulates the Abstraction-Filtration-Comparison (AFC) test to enable interpretable risk evaluation; (2) we introduce a prompt-agnostic latent-space reinforcement learning optimization strategy that suppresses memorized infringing patterns at the noise level while preserving semantic fidelity. Our contributions are threefold: (i) the detection module achieves state-of-the-art performance with significantly improved cross-domain generalization and legally grounded interpretability; (ii) the mitigation strategy effectively reduces copyright infringement risk without degrading generation quality; and (iii) the framework establishes a novel compliance paradigm for AI-generated content.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Assessing whether AI-generated images are substantially similar to copyrighted works is a crucial step in resolving copyright disputes. In this paper, we propose CopyJudge, an automated copyright 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 with multi-LVLM debate to assess the likelihood of infringement and provide detailed judgment rationales. Based on the 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. Besides, our approach can be enhanced by exploring non-infringing noise vectors within the diffusion latent space via reinforcement learning, even without modifying the original prompts. Experimental results show that our 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 could more effectively mitigate memorization and IP infringement without losing non-infringing expressions.

Problem

• Automated copyright infringement identification
• Substantial similarity assessment in AI-generated images
• Mitigation strategy for copyright infringement in AI models

Innovation

• LVLM-based infringement identification
• Abstraction-filtration-comparison test
• Reinforcement learning for mitigation