AI training data and copyright: what have the lawsuits and settlements decided so far?
Courts have split. A US judge found Meta's training on books was fair use (2025), the Third Circuit rejected fair use for a competing AI tool (30 Sep 2026), a German court found OpenAI infringed song lyrics (Nov 2025), and Anthropic settled a books case for $1.5 billion, finally approved 20 Jul 2026. This is general education, not legal advice.
What could your company's records earn? Get a range in 30 seconds. Free, no obligation.
Which AI copyright cases matter most?
Six public cases show where the law stands. Statuses are as of 21 Nov 2026 from the sources listed. Cases move fast, so check the court docket before you rely on any row.
| Case | Court | Status | Date | Source |
|---|---|---|---|---|
| Bartz v. Anthropic | N.D. Cal. | Settled, $1.5 billion; final approval granted | 20 Jul 2026 | Authors Guild |
| Kadrey v. Meta | N.D. Cal. | Partial summary judgment for Meta on fair use | 2025 | US Copyright Office summary |
| Thomson Reuters v. Ross Intelligence | 3rd Cir. | Affirmed: not fair use | 30 Sep 2026 | Justia |
| New York Times v. Microsoft and OpenAI | S.D.N.Y. (consolidated case 25-md-03143) | Summary judgment briefing; new deadlines set | 12 Nov 2026 | Court order, Justia dockets |
| Getty Images v. Stability AI | High Court of England and Wales | Judgment: Getty lost its secondary infringement claim | 4 Nov 2025 | Judiciary UK |
| GEMA v. OpenAI | Munich court | Judgment for the music rights group; undisclosed damages | 11 Nov 2025 | The Guardian |
Appeals after the 2025 rulings were not checked for this page.
What did the Anthropic settlement decide?
On 20 July 2026 Judge Araceli Martínez-Olguín granted final approval of the $1.5 billion class settlement in Bartz v. Anthropic. The court noted the award of about $3,000 per work, which it called four times the $750 statutory minimum for ordinary infringement. Class members release only claims about Anthropic's past acquisition and copying of their works through 25 Aug 2025. Claims over AI outputs and future conduct are kept. The court cut class counsel's fees to about 6.8% of the fund, or about $101.56 million (Authors Guild, 21 Jul 2026).
Has any court said AI training is fair use?
One has, in part. In Kadrey v. Meta, the court called Meta's use of books to train Llama "highly transformative", reasoning that training aims to build a tool that generates varied text while the books exist to be read (US Copyright Office case summary). The books came from shadow libraries. The court held that, absent evidence of market dilution, the factors favored fair use on balance and Meta was entitled to summary judgment.
Has any court said it is not?
Yes, for a competing product. The Third Circuit held on 30 Sep 2026 that ROSS's copying of Westlaw headnotes to train a legal search tool was not fair use, citing its commercial nature, minimal transformative purpose, substantial copying and harm to the market for the original (Justia summary of No. 25-2153). The district court had found 2,243 headnotes protected by copyright.
What did courts outside the US decide?
In England, the High Court found that Stable Diffusion does not store or reproduce Getty's works, and Getty had dropped its primary infringement claims because it lacked evidence that training happened in the UK (Osborne Clarke, 6 Nov 2025). In Germany, the Munich court held that ChatGPT violated German copyright law by using song lyrics to train its models (The Guardian, 11 Nov 2025).
Why does this matter for a company licensing its own data?
Courts put a price on unlicensed material, and AI buyers now want records they can license cleanly. A $1.5 billion settlement is one reason (AI data licensing explained). Your records carry less copyright risk when your company owns them, but buyers still check provenance, as the EU AI Act requires model providers to publish a training-content summary and keep a copyright policy (Article 53, AI Act text). See what AI data buyers check and is it legal to sell company data.
Questions
Is AI training fair use?
No court has settled it for all cases. Rulings so far depend on how the data was obtained and whether the AI tool competes with the original.
Can I sue if my work was used to train a model?
Possibly. Authors in the Anthropic case kept the right to sue over AI outputs and future conduct. Ask a copyright lawyer.
Does a lawsuit affect my company's data license?
Not directly. Your records are yours to license. Buyers will ask you to warrant that you own them.
Sources
- The Authors Guild, 21 Jul 2026: https://authorsguild.org/news/court-grants-final-approval-anthropic-copyright-settlement/
- U.S. Copyright Office, fair use case summary, Kadrey v. Meta, 788 F. Supp. 3d 1026 (N.D. Cal. 2025): https://www.copyright.gov/fair-use/summaries/Kadrey-v-Meta-Platforms-Inc-788-F-Supp-3d-1028-ND-Cal-2025.pdf
- Justia, Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc., No. 25-2153 (3d Cir. 30 Sep 2026): https://law.justia.com/cases/federal/appellate-courts/ca3/25-2153/25-2153-2026-09-30.html
- Justia dockets, order of 12 Nov 2026, New York Times Co. v. Microsoft Corp.: https://docs.justia.com/cases/federal/district-courts/new-york/nysdce/1:2023cv11195/612697/1848
- Courts and Tribunals Judiciary (UK), Getty Images v Stability AI [2025] EWHC 2863 (Ch), 4 Nov 2025: https://www.judiciary.uk/judgments/getty-images-v-stability-ai/
- Osborne Clarke, 6 Nov 2025: https://www.osborneclarke.com/insights/getty-v-stability-ai-stability-ai-generates-big-win-english-courts-landmark-first-judgment
- The Guardian, 11 Nov 2025: https://www.theguardian.com/technology/2025/nov/11/chatgpt-violated-copyright-laws-german-court-rules
- EU AI Act, Article 53: https://artificialintelligenceact.eu/article/53/