Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unlocking Fair Use in the Generative AI Supply Chain: A Systematized Literature Review

Amruta Mahuli, Asia Biega

TL;DR

This paper addresses whether fair use can support training GenAI models and how different stakeholders along the GenAI supply chain value contributions. It employs a PRISMA-based systematic literature review of 20 peer-reviewed studies from 2022–2024, mapped to stakeholder roles, and uses qualitative coding to extract eight value themes. The findings show that certain attributes (creatives' expertise, ideation, implementation; UX-derived content) align with copyright protections, while relational attributes (interpersonal community, lived experience, user-need mappings) are largely unprotected. The authors argue that, due to issues like AI mimicry and overfitting, current fair-use arguments for training data remain contested, and they advocate computational and governance solutions to align copyright rules with the spirit of promoting science and the arts.

Abstract

Through a systematization of generative AI (GenAI) stakeholder goals and expectations, this work seeks to uncover what value different stakeholders see in their contributions to the GenAI supply line. This valuation enables us to understand whether fair use advocated by GenAI companies to train model progresses the copyright law objective of promoting science and arts. While assessing the validity and efficacy of the fair use argument, we uncover research gaps and potential avenues for future works for researchers and policymakers to address.

Unlocking Fair Use in the Generative AI Supply Chain: A Systematized Literature Review

TL;DR

This paper addresses whether fair use can support training GenAI models and how different stakeholders along the GenAI supply chain value contributions. It employs a PRISMA-based systematic literature review of 20 peer-reviewed studies from 2022–2024, mapped to stakeholder roles, and uses qualitative coding to extract eight value themes. The findings show that certain attributes (creatives' expertise, ideation, implementation; UX-derived content) align with copyright protections, while relational attributes (interpersonal community, lived experience, user-need mappings) are largely unprotected. The authors argue that, due to issues like AI mimicry and overfitting, current fair-use arguments for training data remain contested, and they advocate computational and governance solutions to align copyright rules with the spirit of promoting science and the arts.

Abstract

Through a systematization of generative AI (GenAI) stakeholder goals and expectations, this work seeks to uncover what value different stakeholders see in their contributions to the GenAI supply line. This valuation enables us to understand whether fair use advocated by GenAI companies to train model progresses the copyright law objective of promoting science and arts. While assessing the validity and efficacy of the fair use argument, we uncover research gaps and potential avenues for future works for researchers and policymakers to address.
Paper Structure (13 sections, 3 figures, 1 table)

This paper contains 13 sections, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Data Collection
  4. Data Analysis
  5. Results
  6. Value added by creatives while creating expressive work
  7. Value added by ML practitioners while developing ML models
  8. Value added by UX practitioner while developing ML models’ interface
  9. Value derived by users from GenAI outputs:
  10. Analysis and Discussion
  11. Does copyright law protect the attributes valued by stakeholders in GenAI supply chain?
  12. Does the advocacy of fair use doctrine enable the progress of copyright law’s objectives?
  13. Conclusion

Figures (3)

  • Figure 1: PRISMA flowchart of the article selection process
  • Figure 2: Mapping research questions with codes and themes in the paper
  • Figure 3: Mapping copyright law values with valued attributes of stakeholders