Rights & provenance
Training-data copyright lawsuits: the landscape
Last updated 2026-07-22
A running map of the major cases testing whether training AI on copyrighted material without a licence is infringement or fair use.
Why it matters to a buyer
The outcomes directly price the risk of unlicensed data. Buyers need an accurate read, not headlines.
Why it matters to a data owner
The litigation is why licensed supply is worth more — it is the alternative to legal exposure.
Current legal status
The picture is nuanced and case-specific. Thomson Reuters v. Ross (Feb 2025) was the first US ruling to reject a fair-use defense for training data — but it is a non-generative case, so its reach to generative LLMs should not be overstated; the Third Circuit heard the first appellate argument on the question on June 11, 2026, and its decision is pending. Two June 2025 rulings in the Northern District of California went the other way on the training step: Bartz v. Anthropic held training on lawfully acquired books fair use while keeping a pirated library was not, and Kadrey v. Meta granted Meta summary judgment because those plaintiffs failed to show market harm — while cautioning that a better-evidenced case could come out differently. The Bartz claims then settled for $1.5 billion — roughly $3,000 per book — with final approval on July 20, 2026; a settlement sets no precedent. Getty v. Stability in the UK High Court (Nov 2025) was largely a loss for Getty, which withdrew its primary training claims because the training happened outside the UK and lost the secondary-infringement claim, retaining only a narrow trademark win — it is not a rights-holder win on training and decided nothing about whether training infringes. Other cases remain live, including NYT v. OpenAI, the music publishers’ suits against Anthropic, and the Disney, Universal, and Warner Bros. Discovery suits against Midjourney.
What fiund does about it
fiund exists so buyers never have to bet a model on how these cases resolve.
Sources
- Authors Alliance — Bartz v. Anthropic settlement receives final approval (July 2026)
- LawNext — Third Circuit presses Ross and Thomson Reuters on fair use and market harm
- Latham & Watkins — Getty v. Stability: English High Court rejects secondary copyright claim
- CourtListener — Disney Enterprises v. Midjourney docket (C.D. Cal.)
Frequently asked questions
Has any US court ruled that AI training is fair use?
The rulings so far are case-specific, not a blanket answer. In June 2025, Bartz v. Anthropic held training on lawfully acquired books fair use while keeping a pirated library was not, and Kadrey v. Meta granted Meta summary judgment because those plaintiffs failed to show market harm. Thomson Reuters v. Ross (Feb 2025) went the other way for a non-generative tool. The first appellate decision is pending in the Third Circuit.
What did the Bartz v. Anthropic settlement resolve?
The Bartz claims settled for $1.5 billion — roughly $3,000 per book — with final approval on July 20, 2026. A settlement sets no precedent, so it resolved those claims without answering the fair-use question for anyone else.
Did Getty v. Stability decide that AI training is lawful in the UK?
No. Getty withdrew its primary training claims because the training happened outside the UK, lost on secondary infringement, and retained only a narrow trademark win. The judgment decided nothing about whether training itself infringes.
Which major cases are still undecided?
NYT v. OpenAI, the music publishers’ suits against Anthropic, and the Disney, Universal, and Warner Bros. Discovery suits against Midjourney remain live — and the Third Circuit’s decision in Thomson Reuters v. Ross, argued June 11, 2026, is pending.
More on rights & provenance
Want data that clears this in diligence?
Whether you're building a model or sitting on an archive, the first conversation is short and specific.
Send a brief