Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-24 · high
Courts haven't settled whether training on copyrighted books is fair use, and most published authors are already in the corpus without consent.
Why now: TechCrunch published its breakdown of the fair-use fight on August 23.
TechCrunch finds no settled answer to whether training AI models on copyrighted books counts as fair use, laying out the competing legal readings without picking a winner in its August 23 explainer. Most published authors are already in the training corpus, contributing to the tools now competing with them, without knowledge or consent.
That's the part with teeth. The authors had no say. Anyone shipping on top of frontier models is building on data whose legal status is still open, and that exposure doesn't sit only with the labs that did the training.
You can't un-train a model, and you can't swap its foundation under a shipped product without rebuilding on top of a different one. That's what makes this different from an ordinary vendor risk. The liability question sits underneath products that already have users, not in some future release.
Watch for the first ruling that actually decides the fair-use question instead of leaving it open. Until then, every product built on a frontier model is carrying legal exposure nobody's priced yet.
Each link below shares sources, entities, or timing with this story.
Shared entities / Same source domain / Shared topic / Earlier coverage
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); overlapping topics (august, consent).
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); overlapping topics (anyone, august).
Shared entities / Same source domain / Earlier coverage / Tension
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); earlier August coverage from 2026-08-23.
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); earlier August coverage from 2026-08-17.
Shared entities / Same source domain / What happened next
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); picks up the August thread on 2026-08-25.
Shared entities / Same source domain / Earlier coverage
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); earlier August coverage from 2026-08-20.
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); earlier August coverage from 2026-08-13.
Both cover August, TechCrunch; reported by the same outlet (techcrunch.com); earlier August coverage from 2026-08-13.