Case commentary · Independent analysis · 18 November 2026

Bartz v. Anthropic: Fair Use and Training A.I. on Books

Where Your Training Data Comes From Matters

A case commentary on the Northern District of California’s fair use ruling on training AI models with books, and the settlement that followed.

Educational purposes only, not legal advice. Independent commentary by a non-attorney; not filed with any court. Read the court's original opinion for the authoritative text.

Case: Bartz v. Anthropic PBC, No. 3:24-cv-05417-WHA, Order on Fair Use, slip op. (N.D. Cal. June 23, 2025) (Alsup, J.)

fair useAI trainingcopyrightbooksdata provenance

In *Bartz v. Anthropic*, a federal judge held that training large language models on lawfully acquired books was fair use, but that building a permanent central library from pirated copies was not. The class claims later settled, with final approval granted in July 2026. The commentary explains how the court separated acquisition from use, and what the ruling did not decide.

What you will take from it

  • Understand the split holding in plain terms: training was "quintessentially transformative", scanning purchased print books was fair use, and pirated library copies were not.
  • See why the court treated how copies were obtained and kept as a separate question from how they were used.
  • Follow the verified later history: class certification, the denied stay, preliminary approval and the final approval order with a $1.5 billion non-reversionary fund.
  • Note the limits: the authors challenged inputs, not outputs, and there is no appellate ruling in the case.
  • Take general governance observations on data provenance, retention and output monitoring.

For organisations developing or using AI tools, the most durable lesson of *Bartz* may be that the source of training material can matter as much as the use made of it.

From the Open Questions, *Bartz v. Anthropic* case commentary.

Background

In *Bartz v. Anthropic*, a federal judge held that training large language models on lawfully acquired books was fair use, but that building a permanent central library from pirated copies was not. The class claims later settled, with final approval granted in July 2026. The commentary explains how the court separated acquisition from use, and what the ruling did not decide.

Understand the split holding in plain terms: training was "quintessentially transformative", scanning purchased print books was fair use, and pirated library copies were not. See why the court treated how copies were obtained and kept as a separate question from how they were used. Follow the verified later history: class certification, the denied stay, preliminary approval and the final approval order with a $1.5 billion non-reversionary fund.

Note the limits: the authors challenged inputs, not outputs, and there is no appellate ruling in the case. Take general governance observations on data provenance, retention and output monitoring.

Frequently asked questions

What did the court decide?
On summary judgment, it held that using books to train Anthropic’s models was fair use and that digitising purchased print books was fair use, but that downloading pirated copies to build a permanent central library was not.
Is the case still going?
The commentary reports that the class claims settled and that final approval, final judgment and dismissal with prejudice were entered on 20 July 2026. No appellate ruling in the case was found.
Is this legal advice?
No. It is independent commentary by a non-attorney for educational purposes, with general governance observations, not advice about any particular use of data.

Source

More from Thomas Guerriero

See all research at Research, and the main hub at ThomasGuerriero.com. Also: ThomasAGuerrieroPhD.com · ThomasAGuerriero.com.

For educational purposes only; not legal advice. This page summarises independent commentary on a public court opinion and links to the original. A PDF of the commentary is available on this page.

Pursue greatness

Don't get left behind. Chase greatness.

If you want help growing your business, team, or just want to pursue greatness yourself, book a free 15 minute call at the link below.

cal.com/thomas-guerriero-gvx0zs/15min