Nine Ways to Break Copyright Law and Why Our LLM Won't: A Fair Use Aligned Generation Framework

2025-05-25

AI Summary

Large language models (LLMs) risk copyright infringement due to excessive reuse of protected material, and existing refusal-based safeguards severely compromise practical utility. Method: We propose Fair Use-Aligned LLMs (FUA-LLMs), introducing FairUseDB—a first-of-its-kind, expert-annotated dataset covering nine canonical infringement scenarios (18,000 samples)—and a legal-principle-driven generation alignment paradigm. We further design two novel evaluation metrics: Compliance-Aware Utility (CAH) and Weighted Penalty Utility, and employ Direct Preference Optimization (DPO) for fine-grained, copyright-aware fine-tuning. Contribution/Results: Experiments show FUA-LLMs reduce infringing outputs by 20% over state-of-the-art methods. Expert evaluations confirm substantial improvements in legal compliance while preserving high practical utility—demonstrating the effectiveness of legally grounded, utility-preserving alignment.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Large language models (LLMs) commonly risk copyright infringement by reproducing protected content verbatim or with insufficient transformative modifications, posing significant ethical, legal, and practical concerns. Current inference-time safeguards predominantly rely on restrictive refusal-based filters, often compromising the practical utility of these models. To address this, we collaborated closely with intellectual property experts to develop FUA-LLM (Fair Use Aligned Language Models), a legally-grounded framework explicitly designed to align LLM outputs with fair-use doctrine. Central to our method is FairUseDB, a carefully constructed dataset containing 18,000 expert-validated examples covering nine realistic infringement scenarios. Leveraging this dataset, we apply Direct Preference Optimization (DPO) to fine-tune open-source LLMs, encouraging them to produce legally compliant and practically useful alternatives rather than resorting to blunt refusal. Recognizing the shortcomings of traditional evaluation metrics, we propose new measures: Weighted Penalty Utility and Compliance Aware Harmonic Mean (CAH) to balance infringement risk against response utility. Extensive quantitative experiments coupled with expert evaluations confirm that FUA-LLM substantially reduces problematic outputs (up to 20%) compared to state-of-the-art approaches, while preserving real-world usability.

Problem

• Prevent LLMs from reproducing copyrighted content verbatim
• Replace refusal-based filters with fair-use aligned outputs
• Balance legal compliance and response utility in LLMs

Innovation

• Fair Use Aligned Language Models framework
• Direct Preference Optimization fine-tuning
• Compliance Aware Harmonic Mean metrics