# Should AI Companies Pay Royalties for Training Data Scraped from the Web?

> Explore whether scraping human books, journalism, and art is fair use or unprecedented intellectual property theft requiring mandatory licensing fees.

- Category: Artificial Intelligence
- Difficulty: hard
- Suitable for: College, Adults
- Web page: https://debatetopicideas.com/debate-topics/ai-training-data-copyright-compensation/

## Arguments for

- **Prevents the wholesale corporate theft of human creative labor**: Trillion-dollar tech firms scrape journalists' reporting, authors' novels, and artists' portfolios without consent or compensation to build competing commercial products.
- **Directly destroys the economic livelihoods of human creators**: Illustrators, voice actors, and freelance writers find themselves replaced by AI models that were trained directly on their personal copyrighted portfolios.
- **Fair use was never intended to subsidize commercial competitors**: Reading a book to learn is fair use; ingesting millions of books to build an automated synthetic content machine that replaces authors is commercial copyright infringement.
- **Collective licensing models already exist and work smoothly**: The music industry uses ASCAP and BMI to collect and distribute micro-royalties for radio and streaming plays; AI training can adopt identical licensing boards.

## Arguments against

- **Transformative training on public data is the definition of fair use**: AI models do not copy and store text; they learn statistical patterns, grammar, and concepts just like a human student reads hundreds of books to learn to write.
- **Mandatory licensing fees will kill AI competition and empower Big Tech**: Only Microsoft, Google, and Apple have the billions required to license the global internet, permanently outlawing university researchers and open-source startups.
- **Scraping public web data is the foundation of the entire internet**: Google Search, internet archives, and translation engines exist entirely because web scraping of public facts and text is legally protected.
- **Tracking micro-contributions across 10 trillion tokens is impossible**: When an LLM produces a paragraph, attributing fractions of a cent across billions of training sources is a cryptographic and legal impossibility.

## Counter-questions for cross-examination

- Why did OpenAI and Apple sign multi-million dollar licensing deals with publishers like News Corp and Reddit if training is strictly 'fair use'?
- If a human artist studies Picasso's style and paints in that style, why is that legal while AI training on Picasso is questioned?
- Should websites have a universally enforceable 'robots.txt' legal standard that guarantees models cannot train on their text?
- How will AI models avoid 'model collapse' if human artists stop publishing creative work online due to theft concerns?
- Does allowing free training on public data benefit society more than protecting individual copyright holders?

## Research sources

- <https://www.nytimes.com/2023/12/27/business/media/new-york-times-open-ai-microsoft-lawsuit.html>
- <https://hls.harvard.edu/today/the-fair-use-ai-copyright-debate/>

## Related debate topics

- [Should Purely AI-Generated Art and Content Be Eligible for Copyright?](https://debatetopicideas.com/debate-topics/ai-generated-art-copyright-eligibility.md)
- [Should Copyright Protection Be Drastically Reduced to 20 or 25 Years?](https://debatetopicideas.com/debate-topics/drastic-reduction-copyright-length.md)
- [Should Governments Break Up Big Tech Monopolies Like Google and Amazon?](https://debatetopicideas.com/debate-topics/big-tech-monopolies-breakup.md)
- [Should Governments Strictly Regulate Artificial Intelligence?](https://debatetopicideas.com/debate-topics/artificial-intelligence-regulation.md)
- [Should Fully Autonomous Lethal Weapons Be Banned by International Treaty?](https://debatetopicideas.com/debate-topics/autonomous-weapons-ban.md)

---

Source: Debate Topics (https://debatetopicideas.com/). Free to use for classroom and tournament practice.
