ISACL: Internal State Analyzer for Copyrighted Training Data Leakage
This addresses copyright compliance and data privacy issues for AI developers and organizations using LLMs, though it is incremental as it builds on existing detection approaches by shifting to pre-generation analysis.
The study tackled the problem of large language models inadvertently exposing copyrighted training data by introducing a proactive method that analyzes internal states before text generation to detect potential leaks, resulting in effective mitigation of leakage risks with a scalable solution integrated into AI workflows.
Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) but pose risks of inadvertently exposing copyrighted or proprietary data, especially when such data is used for training but not intended for distribution. Traditional methods address these leaks only after content is generated, which can lead to the exposure of sensitive information. This study introduces a proactive approach: examining LLMs' internal states before text generation to detect potential leaks. By using a curated dataset of copyrighted materials, we trained a neural network classifier to identify risks, allowing for early intervention by stopping the generation process or altering outputs to prevent disclosure. Integrated with a Retrieval-Augmented Generation (RAG) system, this framework ensures adherence to copyright and licensing requirements while enhancing data privacy and ethical standards. Our results show that analyzing internal states effectively mitigates the risk of copyrighted data leakage, offering a scalable solution that fits smoothly into AI workflows, ensuring compliance with copyright regulations while maintaining high-quality text generation. The implementation is available on GitHub.\footnote{https://github.com/changhu73/Internal_states_leakage}