Study Reveals AI Models' Copyright Infringement

CosmicTaco · 2025-04-07T11:14:08.092973+05:30

- A new study suggests OpenAI's models have 'memorized' copyrighted content. - Authors, programmers, and rights-holders have filed lawsuits against OpenAI for using their works without permission. - Researchers from University of Washington, University of Copenhagen, and Stanford co-authored the study. - The method involves identifying 'high-surprisal' words in texts to detect memorization. - Findings indicate GPT-4 memorized portions of popular fiction books and New York Times articles. Source: [TechCrunch](https://techcrunch.com/2025/04/04/openais-models-memorized-copyrighted-content-new-study-suggests/)

A new study suggests OpenAI's models have 'memorized' copyrighted content.
Authors, programmers, and rights-holders have filed lawsuits against OpenAI for using their works without permission.
Researchers from University of Washington, University of Copenhagen, and Stanford co-authored the study.
The method involves identifying 'high-surprisal' words in texts to detect memorization.
Findings indicate GPT-4 memorized portions of popular fiction books and New York Times articles.

Source: TechCrunch