Market Design for AI Training Data: Beyond the Copyright Binary

Summary

This academic paper analyzes market design challenges for human-generated content used in AI training. It critiques two polar approaches: a "free-for-all" model (fair use) that fails to compensate creators, and a "strong intellectual property rights" model that, through static Stackelberg game modeling, also underpowers creative incentives—especially for innovative creators (termed the "originality penalty"). A dynamic model reveals another failure: even initially good AI models cause humans to rely more on AI-assisted creation, leading to homogenized content that degrades future model performance (the "curse of precision"). The authors propose a market design featuring a data intermediary that internalizes cross-creator externalities and subsidizes innovative contributions to restore efficiency.