Scaling Laws AI Copyright Lawsuits with Pam Samuelson
Sep 16, 2025
Pam Samuelson, the Richard M. Sherman Distinguished Professor of Law at UC Berkeley, specializes in copyright law and AI's legal implications. She discusses recent court rulings like Bartz v. Anthropic, probing whether training AI on copyrighted material constitutes fair use. The conversation highlights the balance between protecting creators' rights and promoting innovation, while also exploring the transformative nature of AI outputs. Key cases like Warhol vs. Goldsmith are examined for their impact on copyright law, making this a must-listen for anyone interested in the future of intellectual property.
AI Snips
Chapters
Transcript
Episode notes
Alsup Found Training Models Transformative
- Judge Alsup held that using copyrighted works to train generative models is highly transformative and thus fair use.
- He viewed model-building as a non-expressive, transformative market that authors cannot control.
Transformation Protects New Uses But Not Verbatim Copying
- Transformative use protects downstream activities like lectures or models so long as they don't recite substantial expression.
- Direct verbatim recitation from a work can still raise infringement despite transformative learning.
Comparative Law: EU Text-and-Data-Mining Context
- Alsup also allowed scanned copies of books for training as transformative, aligning with EU text-and-data-mining exceptions.
- European law explicitly permits text-and-data-mining for lawful acquisitions, influencing the debate.