Kadrey v. Meta

Richard Kadrey, Sarah Silverman, Christopher Golden, and others v. Meta Platforms, Inc. is a class-action copyright suit by authors alleging that Meta copied their books to train its LLaMA family of large language models. In June 2025 Judge Vince Chhabria granted Meta partial summary judgment on fair use, but on grounds the opinion described as narrow: the authors lost because they failed to develop the right evidentiary record, and the ruling expressly did not hold that training on copyrighted material is categorically lawful. It is one of the two early training-data fair-use decisions, alongside Bartz v. Anthropic.

Background

The plaintiffs allege that Meta copied their books — sourced from the Books3 dataset and shadow libraries — to train its LLaMA family of models. The suit was filed in July 2023 in the Northern District of California. Later amended complaints added further author-plaintiffs, including Ta-Nehisi Coates.

Claims

The authors brought copyright-infringement claims over the reproduction of their books in Meta's training corpus, and a claim under the DMCA for removal of copyright-management information (CMI) from those works.

The fair-use ruling (June 2025)

In June 2025 Judge Chhabria denied the plaintiffs' motion for partial summary judgment and granted Meta's cross-motion for partial summary judgment on fair use. He also granted Meta summary judgment on the DMCA copyright-management-information claim, reasoning that because the copying was not infringement, removing CMI could not have furthered infringement.

The opinion framed the outcome as a win for Meta on narrow grounds, and limited what it decided in three respects. The authors lost because they "failed to present meaningful evidence" on the effect of training LLMs on the market for their books, rather than because training is categorically fair use. Chhabria stated that the ruling does not stand for the proposition that Meta's use of copyrighted material to train its models is lawful — only that these plaintiffs made the wrong arguments and failed to develop the right record. He raised a "market dilution" theory: that flooding the market with AI-generated competing works could be the cognizable harm, a theory the plaintiffs did not adequately pursue and one that he suggested could favor future plaintiffs who do.

Subsequent procedural history (2026)

The June 25, 2025 order resolved the training-use fair-use question but not the case. On March 25, 2026, Judge Chhabria granted the plaintiffs leave to amend their complaint to add contributory-infringement and uploading-based (distribution) claims arising from Meta's use of shadow-library sources, keeping those theories in the district court (Source: cdn.arstechnica.net).

In July 2026, Chhabria denied the authors' motion to certify for interlocutory appeal to the Ninth Circuit the portion of the summary-judgment ruling concerning the initial downloading of the books, reasoning that a final judgment would soon be entered in at least one of the Meta cases before the court, allowing the downloading issue and other questions about the use of copyrighted materials in AI training to be presented to the court of appeals together (Source: chatgptiseatingtheworld.com). As of July 2026 the case remains active in the district court on the amended claims (docket: courtlistener.com).

Relation to other rulings

The "market dilution" framing has been read as indicating how future plaintiffs might prevail, leaving the fair-use question for training data unsettled rather than resolved in developers' favor.

Bartz v. Anthropic (the Alsup ruling, also 2025) and Kadrey (the Chhabria ruling) are the two early training-data fair-use decisions. Both went the AI developer's way on the training-use question, but on different reasoning. Bartz separately produced a settlement of approximately $1.5B over how the books were acquired (piracy), a question distinct from fair use. Both rulings serve as the baseline against which the 105+ active US AI copyright suits compiled in AI Copyright Litigation — Analysis are measured, including the later publisher class action in Hachette et al. v. Meta (and Mark Zuckerberg), where Kadrey is the Meta-side counterpart to Bartz. Kadrey is also cited as the operative Meta precedent on training-data fair use in Training AI is Not Fair Use? — Ben Thompson (Stratechery, July 2025).

Training-data transparency law — California AB 2013 — Generative AI Training Data Transparency — requires developers to document the IP status of their corpora; such disclosures can become evidence in this kind of litigation.

Relationships

• litigates: Meta AI.
• related: Bartz v. Anthropic — sibling 2025 training-data fair-use ruling.
• related: Hachette et al. v. Meta (and Mark Zuckerberg) — later publisher class action against Meta; measured against the post-Bartz/Kadrey baseline.
• related: AI Copyright Litigation — Analysis — umbrella tracker.
• related: Training AI is Not Fair Use? — Ben Thompson (Stratechery, July 2025), California AB 2013 — Generative AI Training Data Transparency.

Sources

• (Source: bakerlaw.com) — BakerHostetler AI litigation case tracker: ruling summary, fair-use and DMCA dispositions.
• (Source: law.justia.com) — Justia docket entry: order denying plaintiffs' motion for partial summary judgment and granting Meta's cross-motion.