November 4, 2025
Fair Use or Fair Uproar?
AI and Copyright: Expanding Copyright Hurts Everyone–Here's What to Do Instead
EFF says let AI read the web; commenters call it a creative heist
TLDR: EFF says AI training should be covered by fair use to protect science and competition, not locked behind licenses. The comments erupt: artists cry “theft,” skeptics accuse EFF of cheerleading Big Tech, and the debate splits between open research and protecting creative livelihoods—stakes that affect everyone who makes or uses content.
The EFF lit a match with a fiery take: you shouldn’t need a “permission slip” to read a webpage—humans or software. They argue fair use (a legal exception that lets you use small bits of copyrighted stuff without permission) should shield AI training and TDM (text and data mining) because it powers science like AlphaFold and space research, and licensing everything would hand the keys to a few giants. They cite Thomson Reuters v. Ross Intelligence as a cautionary tale, and even the FTC warning about data monopolies.
The comments? Absolute chaos. One camp calls it “GenAI agitprop” and blasts the EFF for dodging the art crisis: models trained on artists’ work, then sold back to the world. Another asks how “fair use” magically covers scraping the entire internet. A standout meme: AI as a “digital combine harvester” mowing down the commons. Music folks go full dystopia: AI scores everywhere, composers sidelined, Big Tech as the only buyer.
Supporters are quieter but insist: science needs data, and making researchers negotiate with millions of rights holders is impossible. The thread devolves into punchlines: “Let AI read? Sure. Let AI sell my style? Pay up.” and “EFF = Everyone’s Files Free?” The result: a high-drama clash over who gets to read, remix, and profit from the world’s creativity.
Key Points
- •The article argues that requiring licenses for AI training materials would hinder innovation, expression, and research.
- •It claims fair use protects ML and TDM research, and licensing vast datasets is cost-prohibitive and impractical.
- •Empirical evidence is cited that TDM research is more prevalent in countries with protections against copyright control.
- •AlphaFold is presented as an open-source deep learning tool enabling broad scientific advances.
- •The piece warns that licensing requirements would limit competition and cites the FTC and the Thomson Reuters v. Ross Intelligence case as examples.