What Is the AI Copyright Lawsuit?
The AI copyright lawsuit is a group of separate court cases over whether AI companies may train models on copyrighted work, keep pirated copies, and sell products that reproduce it, not one class action or one MDL. The cases sit in different courts, move at different speeds, and have produced different answers, so no single ruling decides the whole question.
The plaintiffs fall into five groups: authors, news publishers, music publishers and record labels, visual artists, and film studios. The defendants include OpenAI, Microsoft, Anthropic, Meta, Google, Midjourney, Stability AI, Suno and Udio. Judges have reached different conclusions on fair use depending on the evidence each plaintiff built, which is why trackers count dozens of cases but only a few merits rulings.
“The OpenAI Defendants move under 28 U.S.C. § 1407 to centralize this litigation in the Northern District of California. This litigation consists of twelve actions pending in the Northern District of California and the Southern District of New York, as listed on Schedule A.”
Source: U.S. Judicial Panel on Multidistrict Litigation, In re: OpenAI, Inc., Copyright Infringement Litigation, MDL No. 3143, Transfer Order, Document 85 (April 3, 2025)
Why Are People Filing AI Copyright Lawsuits?
People are filing AI copyright lawsuits because they allege AI companies used their creative work to build commercial products without asking or paying. Authors say their books were copied from shadow libraries, news publishers say their articles were reproduced in chatbot answers, and artists and studios say image generators can recreate their styles and characters.
Statutory damages give the claims real weight. Under 17 U.S.C. § 504(c), damages generally run from $750 to $30,000 per registered work and can reach $150,000 per work if infringement is found willful, so a training set of millions of works creates very large theoretical exposure. Licensing is the other driver: when AI companies pay some rights holders, other rights holders argue that an existing market for AI licensing already exists.
What Injuries Are Included in the AI Copyright Lawsuit?
The AI copyright lawsuit includes four distinct kinds of claimed harm, pled in different combinations across the cases. Each is defined in plain language below, followed by what plaintiffs allege.
Training-Data Copying
Training-data copying is a legal claim that a company made copies of protected works to feed an AI model, without a license from the rights holder.
Authors, news publishers and music publishers allege that OpenAI, Microsoft, Anthropic and Meta copied their works into the datasets used to train large language models. The defendants answer that training is a transformative fair use, and the U.S. courts that have ruled so far have not agreed on how far that answer reaches.
Pirated-Library Acquisition
Pirated-library acquisition is a legal claim that a company downloaded or stored copies of works from unauthorized sources, such as shadow libraries, regardless of what it later did with them.
Several complaints allege that books were taken from sites such as LibGen, Z-Library and OceanofPDF. A December 2025 complaint filed by authors who opted out of the Anthropic class settlement seeks up to roughly $900,000 per book by stacking $150,000 in statutory damages across six defendants.
Infringing Outputs
Infringing outputs is a legal claim that what an AI system produces, such as text, lyrics or images, reproduces or closely imitates a protected work.
News publishers allege that ChatGPT can generate detailed summaries of their articles, music publishers allege that Claude can reproduce lyrics, and film studios allege that Midjourney can generate recognizable characters. In MDL No. 3143, the court allowed the output-based claims against OpenAI to proceed past a motion to dismiss.
Market Dilution
Market dilution is a legal theory that an AI product can flood a market with competing content, harming the value of the original works even when no single output copies them.
The theory comes from the court's own reasoning in Kadrey v. Meta, which described it as potentially stronger than the arguments the thirteen authors actually made. The court found that the thirteen authors had not built a record to support it, and later plaintiffs are shaping their evidence around it.
Who Qualifies for the AI Copyright Lawsuit?
You may qualify if:
- You own a copyright registration for a book, article, song, image or other work that you believe was used to train a commercial AI model
- Your work appears in a dataset or source that a pending case identifies, such as a pirated library named in a complaint
- You opted out of a class settlement and want to know whether to bring an individual claim
You may not qualify if:
- Your work was never registered with the U.S. Copyright Office, which generally must happen before an infringement suit can be filed
- Your work was covered by the Anthropic author settlement and you did not claim by the March 30, 2026 deadline
- The time limit for your claim has likely run, since copyright claims generally carry a three-year limitations period measured from when you knew or should have known
Eligibility depends on your registration status, the specific AI company and model involved, whether a case has a class definition that covers you, and your jurisdiction. Only an attorney experienced in copyright litigation can confirm whether you qualify.
What Is the Current Status of the AI Copyright Lawsuit?
The current status of the AI copyright lawsuit depends on the track. The Anthropic author case has settled, the Thomson Reuters appeal has been decided, the OpenAI MDL is heading toward summary judgment, and the music and visual-art cases are still moving through discovery and pretrial.
Bartz v. Anthropic PBC
Bartz v. Anthropic is the author class action that produced the only large AI copyright settlement so far. Judge William Alsup split the fair use question in a June 23, 2025 summary judgment order, and the parties announced a class settlement on August 25, 2025.
| Proceeding Type | Federal class action |
| Court | U.S. District Court, Northern District of California, No. 3:24-cv-05417 |
| Judge | Honorable William H. Alsup |
| Preliminary Approval | September 25, 2025 |
| Claims Deadline | March 30, 2026 |
| Final Approval | July 20, 2026 |
In re: OpenAI, Inc., Copyright Infringement Litigation, MDL No. 3143
MDL No. 3143 is the federal proceeding that combines authors, news publishers and a YouTube-creator class against OpenAI and Microsoft. Judge Sidney H. Stein is overseeing discovery, which has included orders requiring OpenAI to produce tens of millions of ChatGPT logs.
| Proceeding Type | Federal multidistrict litigation, 28 U.S.C. § 1407 |
| Court | U.S. District Court, Southern District of New York, No. 1:25-md-03143 |
| Judge | Honorable Sidney H. Stein |
| Centralized | April 3, 2025, twelve actions |
| Next Milestone | Summary judgment briefing in the New York Times case, reported to run to November 20, 2026 |
Thomson Reuters v. Ross Intelligence Inc.
Thomson Reuters v. Ross Intelligence is the case that produced the first U.S. ruling rejecting fair use for AI training data. The Third Circuit affirmed Thomson Reuters' win on September 29, 2026, and reports say parts of the opinion's reasoning remain under seal pending redactions.
| Proceeding Type | Federal copyright case, interlocutory appeal decided |
| Court | U.S. District Court, District of Delaware, No. 1:20-cv-00613, appeal to the Third Circuit |
| Trial Judge | Honorable Stephanos Bibas |
| District Ruling | February 11, 2025, 2,243 headnotes found copied |
| Appeal Decision | September 29, 2026, affirmed |
Disney, Universal and Warner Bros. v. Midjourney
The studios' case against Midjourney is the main pending test of whether an image generator is liable for outputs that depict recognizable characters. It is in discovery, with a jury demanded, and sits beside Andersen v. Stability AI, the artist class action that had a first jury trial date of September 8, 2026.
| Proceeding Type | Federal copyright case, jury demanded |
| Court | U.S. District Court, Central District of California, No. 2:25-cv-05275 |
| Judge | Honorable John A. Kronstadt |
| Filed | June 11, 2025 (Disney and Universal), September 4, 2025 (Warner Bros.) |
| Latest Reported Step | Informal discovery conference, May 22, 2026 |
How Much Do AI Copyright Lawsuit Settlements Pay?
AI copyright lawsuit settlements have paid about $3,000 per covered work in the largest settlement so far, the $1.5 billion Bartz v. Anthropic class settlement covering an estimated 500,000 books.
That figure is a per-work average before attorney fees and before the split between authors and publishers, and it applies only to Anthropic's past use of the covered books. No other defendant has reached a comparable settlement, so what any future recovery would be worth depends on the case, the registration status of the works and whether a court certifies a class.
Graeber, one of the three named plaintiffs, told NPR he expects around $3,100 for each of his two books, and that more than $100 million of the fund went to the plaintiffs' lawyers. The settlement is paid in installments, with the final tranche scheduled for September 2027, according to a book-industry tracker's May 2026 snapshot.
Music cases have settled differently, through licensing deals rather than a single fund. Universal Music Group settled with Udio on October 29, 2025, Warner Music Group settled with Udio on November 19, 2025, and Warner also settled with Suno in November 2025, each with licensing terms for a new service.
AI Copyright Lawsuit Updates and Case Status
AI copyright lawsuit updates below follow the major rulings, settlements and filings across every defendant, listed from newest to oldest. Each entry names its own case, so the timeline can be read from any point.
AI Copyright Lawsuit Timeline
Third Circuit Affirms Thomson Reuters' Win Over Ross Intelligence
The Third Circuit affirmed summary judgment for Thomson Reuters in Thomson Reuters v. Ross Intelligence on September 29, 2026, holding that Ross's copying of Westlaw headnotes to train its legal-research tool was not fair use. Reports describe it as the first federal appellate ruling on fair use in AI training and note that the opinion distinguishes Ross's non-generative search tool from large language models.
Andersen v. Stability AI Jury Trial Was Scheduled to Begin
A book-industry tracker listed September 8, 2026 as the first jury trial date in Andersen v. Stability AI, the artists' class action before Judge William H. Orrick in the Northern District of California. This page has not independently confirmed whether the trial began or what the outcome was, so readers should check the docket, No. 3:23-cv-00201.
New York Times and OpenAI File Cross-Motions on Fair Use
The New York Times, OpenAI and Microsoft filed cross-motions for summary judgment on fair use in MDL No. 3143 on or about September 4, 2026, with briefing reported to run to November 20, 2026. One report says a ruling from Judge Stein is not expected until 2027.
Justice Department Files Statement Backing OpenAI on Fair Use
The U.S. Department of Justice filed a statement of interest in MDL No. 3143 on September 1, 2026, arguing that training large language models on copyrighted text is fair use. News publishers asked the court to give the statement no weight, and a statement of interest does not bind the judge.
Judge Approves Anthropic's $1.5 Billion Author Settlement
A federal judge in San Francisco gave final approval to the $1.5 billion Bartz v. Anthropic settlement on July 20, 2026, after more than 91 percent of covered authors and publishers claimed their share. Anthropic's counsel told NPR the company maintains that training AI on books is fair use.
Getty's Trademark Claims Survive Dismissal in California
Judge Trina L. Thompson ruled on April 23, 2026 that Getty Images' direct copyright, trademark and California unfair-competition claims against Stability AI could proceed, while dismissing its false-copyright-information claim without prejudice. The trademark theory rests on Stable Diffusion outputs that reproduced Getty's watermark.
OpenAI MDL Holds Discovery and Settlement Conferences
Judge Stein's MDL No. 3143 held a discovery status conference and a settlement conference between March 10 and 13, 2026. No settlement was announced.
Court Orders OpenAI to Produce Millions More ChatGPT Logs
The court in MDL No. 3143 granted the plaintiffs' motion to compel on March 9, 2026, ordering OpenAI to produce reservoirs of 78 million and 10 million logs on top of the 20 million-log sample already ordered.
Supreme Court Declines Thaler's AI-Authorship Appeal
The U.S. Supreme Court denied certiorari in Thaler v. Perlmutter on March 2, 2026, leaving in place the rule that copyright requires human authorship. The ruling concerns who can own copyright in AI output, not whether training is lawful.
Music Publishers File Second Lawsuit Against Anthropic
Concord, UMG and ABKCO filed a second suit against Anthropic in the Northern District of California on January 29, 2026, built on discovery from the Bartz case that surfaced more than 20,000 lyrics, compositions and sheet-music files. The complaint seeks damages the publishers put in excess of $3 billion.
Judge Stein Affirms Order to Produce 20 Million ChatGPT Logs
Judge Stein ruled on January 5, 2026 that OpenAI must produce the full 20 million-log anonymized sample in the New York Times litigation, rejecting OpenAI's proposal to hand over a hand-picked subset.
Opt-Out Authors Sue Six AI Companies
Authors who opted out of the Anthropic settlement, including John Carreyrou, filed suit on December 22, 2025 against Anthropic, Google, OpenAI, Meta, xAI and Perplexity, alleging books were copied from LibGen, Z-Library and OceanofPDF.
Hamburg Court Rejects Kneschke's Appeal Against LAION
The Higher Regional Court of Hamburg dismissed photographer Robert Kneschke's appeal against LAION on December 10, 2025, holding that LAION's dataset work qualified for the scientific-research exception and that a plain-English opt-out was not machine-readable. Leave to appeal to Germany's Federal Court of Justice was granted.
Warner Music Group Settles With Udio
Warner Music Group announced on November 19, 2025 that it had settled its lawsuit against Udio, following Universal Music Group's October 29 settlement. Sony Music, the third label plaintiff, had not settled as of year-end 2025 reviews.
Munich Court Finds OpenAI Liable in GEMA Lyrics Case
The Munich Regional Court ruled for GEMA against OpenAI on November 11, 2025, holding that memorizing and reproducing nine song lyrics infringed German copyright law and rejecting OpenAI's text-and-data-mining defense.
UK High Court Rules Model Weights Are Not Copies
Mrs Justice Joanna Smith ruled for Stability AI against Getty Images on secondary copyright infringement on November 4, 2025, holding that Stable Diffusion's model weights are not themselves an infringing copy, while finding narrow, historic trademark infringement from watermarks.
Judge Stein Lets OpenAI Output Claims Proceed
Judge Stein held on October 27, 2025 that the MDL plaintiffs had alleged some ChatGPT outputs a reasonable jury could find substantially similar to their works, denying OpenAI's motion to dismiss the output-based claims without ruling on fair use.
Anthropic and Authors Announce $1.5 Billion Settlement
Anthropic and the Bartz plaintiffs announced a $1.5 billion class settlement on August 25, 2025, the largest copyright settlement in U.S. history, with preliminary approval granted on September 25, 2025.
Judge Chhabria Grants Meta Fair Use Win on Narrow Record
Judge Vince Chhabria granted Meta summary judgment on fair use in Kadrey v. Meta on June 25, 2025, but wrote that the ruling covers only thirteen authors and does not mean Meta's training is lawful. Claims over torrent distribution continued, and the authors sought to add a contributory-infringement claim in December 2025.
Judge Alsup Splits the Fair Use Question in Bartz
Judge Alsup ruled on June 23, 2025 that Anthropic's training on lawfully acquired books was fair use but that keeping over seven million pirated copies in a central library was not, sending that claim toward trial.
Copyright Office Says Fair Use Cannot Be Presumed for AI Training
The U.S. Copyright Office released a pre-publication Part 3 report on May 9, 2025 concluding that fair use for AI training must be judged case by case and that lost licensing markets can weigh against it.
Judicial Panel Centralizes Twelve OpenAI Cases in New York
The Judicial Panel on Multidistrict Litigation created MDL No. 3143 on April 3, 2025, transferring twelve actions against OpenAI and Microsoft to Judge Stein in the Southern District of New York for pretrial proceedings.
First U.S. Ruling Rejects Fair Use for AI Training Data
Judge Stephanos Bibas granted partial summary judgment to Thomson Reuters against Ross Intelligence on February 11, 2025, finding Ross's use of Westlaw headnotes to build a competing legal-research tool was not fair use.
How to File an AI Copyright Lawsuit
The 5 steps below explain how to file an AI Copyright lawsuit.
- Confirm your copyright registration: Federal law generally requires a registration before you can sue for infringement, and registration timing affects whether statutory damages are available.
- Check whether your work was in a known dataset: Public lookup tools and the Anthropic settlement portal can show whether a title appeared in sources such as Books3 or LibGen, though a match alone does not prove a claim.
- Review your publishing and licensing contracts: Many pre-2024 contracts are silent on AI training, and some publishers have signed AI licensing deals, so your contract affects who holds the claim.
- Find out whether a pending class already covers you: Class definitions and opt-out rights differ by case, and choosing to opt out of a settlement ends your share of that fund.
- Consult a copyright attorney before acting: An attorney can confirm registration, the limitations period and which defendant and court fit your facts.
Answer a few questions and a case reviewer will follow up. Free, no obligation.
A case reviewer will follow up soon.
What Evidence Do I Need for an AI Copyright Lawsuit?
The strongest evidence in an AI copyright case is proof that your registered work was copied into a specific company's training data or reproduced in its output. That starts with the registration certificate and the date of first publication, since registration timing generally affects statutory damages under 17 U.S.C. § 412.
Evidence of use comes from three places: dataset lookups showing a title in a pirated or scraped source, court discovery in cases such as Bartz, and saved examples of chatbot or image-generator output that closely track your work. Contracts matter as well, because a publisher agreement may transfer or share the right to sue, and only an attorney reviewing your documents can say which evidence is enough.
What Have AI Companies Argued in the AI Copyright Lawsuit?
The AI companies' positions share one core argument, that training a model on copyrighted work is a transformative fair use, but each defendant adds its own defenses. The table summarizes the stated positions as reported in court filings and press coverage.
| Defendant | Stated position | Where it appears |
|---|---|---|
| OpenAI and Microsoft | Fair use, de minimis copying and no substantial similarity in outputs; OpenAI admitted some factual allegations about the datasets used for GPT-2 and GPT-3 | MDL No. 3143 pleadings and motions |
| Anthropic | Training AI on books is fair use, according to its deputy general counsel | Statement to NPR, July 2026 |
| Meta | Training Llama on books is fair use, a position that prevailed on the record in Kadrey | Kadrey v. Meta summary judgment |
| Midjourney | Fair use described as quintessentially transformative, plus license, de minimis use and DMCA safe-harbor defenses | Answer filed August 6, 2025 |
The Justice Department sided with the AI industry on the core question in September 2026 through a statement of interest in MDL No. 3143, a position the news plaintiffs asked the court to disregard.
How Do the AI Copyright Lawsuit Fair Use Rulings Compare?
The fair use rulings in the AI copyright lawsuit do not point in one direction, because each court looked at different facts, defendants and evidence of market harm. Two judges in the same California district split on how much the market-harm factor matters, and courts abroad have applied different laws entirely.
For the OpenAI MDL on its own, see the OpenAI lawsuit page, and for a landscape view centered on those defendants see the AI copyright landscape guide.
| Case | Court and judge | Ruling | Scope |
|---|---|---|---|
| Thomson Reuters v. Ross | D. Del., Judge Bibas; Third Circuit affirmed Sept. 29, 2026 | Fair use rejected, February 11, 2025 | Non-generative legal search tool |
| Bartz v. Anthropic | N.D. Cal., Judge Alsup | Training on lawfully acquired books fair use; pirated central library not, June 23, 2025 | Anthropic's past conduct; case settled |
| Kadrey v. Meta | N.D. Cal., Judge Chhabria | Fair use granted to Meta, June 25, 2025 | Thirteen authors only; market-dilution theory left open |
| Getty Images v. Stability AI (UK) | High Court, Mrs Justice Smith | Model weights not an infringing copy, November 4, 2025 | UK law; narrow watermark trademark finding |
| GEMA v. OpenAI | Munich Regional Court | Memorization and output of lyrics infringed, November 11, 2025 | German law; nine lyrics |
| Kneschke v. LAION | Higher Regional Court of Hamburg | Research exception applied, December 10, 2025 | German law; appeal to Federal Court of Justice allowed |
Because the U.S. rulings split and none has been tested in a final judgment against a generative AI developer on the market-dilution theory, the open question is likely to reach a higher court in the OpenAI MDL or the music and image cases.
Frequently Asked Questions
What is the AI copyright lawsuit about?
The AI copyright lawsuit is about whether AI companies can lawfully train models on copyrighted books, articles, lyrics, images and characters, keep pirated copies, and sell products whose outputs reproduce protected work. It is a group of separate cases in different courts, not one case.
Is the AI copyright lawsuit a class action or an MDL?
The AI copyright lawsuit is both and neither, because the cases use different procedures. The OpenAI and Microsoft cases are centralized in MDL No. 3143 in New York, the Anthropic author case was a certified class action that settled, and the Disney, Universal and Warner Bros. case against Midjourney is a standalone suit.
Has the AI copyright lawsuit been settled?
Parts of the AI copyright lawsuit have settled, but the whole has not. The $1.5 billion Anthropic author case and the Udio and Suno music label cases settled, while the OpenAI MDL, the Midjourney case and the Getty case against Stability AI remain unresolved.
How much will authors get from the Anthropic settlement?
Authors get about $3,000 per work from the Anthropic settlement before fees, split with publishers where a publisher holds rights. The claims deadline was March 30, 2026.
Is training AI on copyrighted books legal under fair use?
Training AI on copyrighted books has been found fair use in two U.S. rulings, Bartz v. Anthropic and Kadrey v. Meta, but neither is a final answer. The Bartz court still found that keeping pirated copies was not fair use, and the Kadrey court said its ruling covers only thirteen authors.
Is the Thomson Reuters v. Ross case the same as the generative AI cases?
Thomson Reuters v. Ross is not the same as the generative AI cases. Ross built a non-generative legal search tool from Westlaw headnotes, and the courts distinguished it from large language models, so its fair use result does not automatically apply to ChatGPT, Claude or Llama.