Matthew Stoller says copyrighted training data is not fair use and licensing could reshape AI

GaryMarcus · x · 2026-07-26

Gary Marcus reposted Matthew Stoller arguing that training on copyrighted materials is not fair use.

The core claim is that if licensing becomes mandatory, it could create new content business models—and that AI systems would still ultimately improve, even under a licensing regime.

Original post →

More from AGI Musings

AGI Musings channel →