A US district judge dismisses most claims in a lawsuit by Sarah Silverman and other authors against Meta's use of copyrighted material to train Llama models
Winston Cho / Hollywood Reporter :
Context & Ripple Effects
This was an early procedural setback in the authors’ broader challenge to AI training practices, filed amid parallel claims that Llama and ChatGPT used infringing training material. It narrowed the dispute without ending the underlying conflict over whether copyrighted works can be used to develop models.
The issue remained live: the related Kadrey case was later allowed to proceed, before a later ruling found Meta’s book-training use fair use on the arguments presented. That sequence makes this dismissal a marker of how strongly case framing and evidence can shape AI-copyright litigation.
First-order effects
- Meta’s immediate exposure in this suit narrows because most of the authors’ claims are out, while the plaintiffs can pursue only the claims that survived.
- The authors lose leverage on the dismissed theories and must concentrate their litigation on the remaining allegations rather than the full original complaint.
Second-order effects
- The decision gives other AI developers a procedural reference point in copyright suits, though it does not establish that training on copyrighted works is lawful; similar claims against other model makers may still turn on their own pleadings and facts.
- Rights holders may refine complaints to better connect alleged copying, model development, and concrete harm—a pressure visible in the later decision allowing the Kadrey authors’ case to continue.
Third-order effects
- AI-training copyright disputes are likely to be resolved through a mix of pleading standards and fact-specific fair-use analysis, rather than a single early ruling that settles the issue for all models and datasets.
- If this pattern persists, developers and rights holders will place more value on documenting training-data practices and on litigation strategies tailored to particular uses; Meta’s later fair-use ruling over Llama book training underscores that outcomes can depend on the record and arguments before the court.
The trend: This is one data point in the gradual legal sorting of generative-AI training disputes, where procedural rulings and fact-specific fair-use decisions are defining the boundaries incrementally.