Blockchain-Based Ecosystem for Fair And Ownership-Preserving Document-Sharing in LLM Use Cases

International Conference on Blockchain Computing and Applications (BCCA), pp. 115–122

Abstract

Large language models (LLMs) have revolutionized natural language processing but face significant challenges in accessing domain-specific knowledge while respecting intellectual property rights. As content providers increasingly restrict access to valuable training data, systems that fairly compensate knowledge contributors while preserving document ownership are urgently needed. This paper presents ChainShare, an innovative blockchain-based ecosystem that creates a trustless framework for incentivized, ownership-preserving document-sharing in retrieval-augmented generation (RAG) pipelines. The system implements a novel architecture where document processing occurs entirely client-sided, combining chunk-level encryption with smart contracts to ensure content providers maintain sovereignty over their intellectual property. Through this design, ChainShare guarantees that knowledge contributors retain complete control over their content until fair compensation is automatically distributed through blockchain-based reward mechanisms. Our evaluation demonstrates ChainShare’s effectiveness in enhancing domain-specific information retrieval compared to traditional LLMs, with result quality metrics improving by up to $63.6 \%$ and blockchain integration being crucial for $72.7 \%$ of participants’ trust in the system. The economic model proves especially promising, with $90.9 \%$ of participants expressing increased willingness to pay for specialized knowledge access after experiencing the system. We have developed a fully functional web application implementing the ChainShare ecosystem that demonstrates these capabilities in practice. This work contributes to ethical AI development by demonstrating how decentralized technologies can simultaneously address challenges in data quality, ownership preservation, and fair compensation for intellectual property.

Authors 2

1. Sascha Nievelstein Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University,Information Systems & Databases,Aachen,Germany

2. Wolfgang Prinz Aachen

   RWTH Aachen University · Fraunhofer Institute for Applied Information Technology

   Affiliation as printed

   RWTH Aachen University,Fraunhofer FIT,Sankt Augustin,Germany