Position: Iterative Online-Offline Joint Optimization is Needed to Manage Complex LLM Copyright Risks

Yanzhou Pan, Jiayi Chen, Jiamin Chen, Zhaozhuo Xu, Denghui Zhang

Abstract: The infringement risks of LLMs have raised significant copyright concerns across different stages of the model lifecycle. While current methods often address these issues separately, this position paper argues that the LLM copyright challenges are inherently connected, and independent optimization of these solutions leads to theoretical bottlenecks. Building on this insight, we further argue that managing LLM copyright risks requires a systemic approach rather than fragmented solutions. In this paper, we analyze the limitations of existing methods in detail and introduce an iterative online-offline joint optimization framework to effectively manage complex LLM copyright risks. We demonstrate that this framework offers a scalable and practical solution to mitigate LLM infringement risks, and also outline new research directions that emerge from this perspective.

Lay Summary:

Large language models (LLMs) can unintentionally reproduce copyrighted content, raising serious legal and ethical concerns. However, current methods to address these risks focus on isolated stages—like training or output filtering—without considering how these stages interact.
We show that the fragmented approach has inherent limitations and propose a unified, joint optimization framework that coordinates online and offline copyright risk controls throughout the LLM lifecycle.
This unified framework enables more effective risk mitigation and better aligns with real-world deployment needs. It not only offers a scalable solution to manage legal exposure but also opens new research directions for building AI systems that are both powerful and compliant with copyright law.

Primary Area: System Risks, Safety, and Government Policy

Keywords: LLM, Copyright

Submission Number: 141