Latent Space AI AI’s Copyright Reckoning
Sep 15, 2025
A recent $1.5 billion settlement sparks discussions about the future of AI and copyright law. The impact of a court ruling permits AI companies to leverage copyrighted material under fair use, raising both ethical and financial questions. Challenges in sourcing original data for AI-generated images are highlighted, showcasing Adobe's efforts to create a fair system for photographers. The potential consequences of these changes could either boost or hinder advancements in the industry.
AI Snips
Chapters
Transcript
Episode notes
Historic $1.5B Copyright Settlement
- Anthropic paid $1.5B to settle claims tied to training on copyrighted books from pirated 'shadow libraries'.
- The payout is the largest in US copyright history and clarifies industry risk around scraped book data.
How Claude Got A Better Writing Tone
- Jaeden describes Anthropic scraping pirated 'shadow libraries' full of books to rapidly improve Claude's writing tone.
- He says they also scanned bought books with robots to transcribe pages into training data.
Fair Use Ruling Shapes Precedent
- Judge William Alsup ruled training on copyrighted material can be fair use if the model transforms the works rather than replicates them.
- The court distinguished lawful training from the separate piracy problem tied to illegally obtained copies.