Restored research edition · Records retain their source dates. Check the original source for current status, deadlines and job openings.
Copyright cases.
Understand the dispute, find what the court decided and read the source. Search selected copyright and AI case events by the questions that matter to your work.
Find a party, compare an issue and open the document behind the decision. Each record describes a particular event; later proceedings may exist.
What did U.S. courts decide about the copying?
Compare the specific use a court assessed. A mixed result means different kinds of copying received different treatment; the chart does not assign a winner to an entire lawsuit.
| Ruling and copying at issue | Fair use accepted | Different results by use | Fair use rejected |
|---|---|---|---|
The Third Circuit affirmed partial summary judgment for Thomson Reuters: the headnotes at issue were copyrightable and ROSS's copying was not fair use. The opinion distinguishes ROSS's non-generative search engine from generative models. Read the ruling’s source | |||
The court granted Meta summary judgment on the authors' training claim on the record presented. The decision did not establish that all LLM training is fair use. Read the ruling’s source | |||
The court treated model training, digitizing lawfully purchased books, and retaining pirated library copies separately. Training and the purchased-book digitization were fair use; the pirated library copies were not. Read the ruling’s source |
This selection includes a ruling with different outcomes for different uses. Open that row before treating “AI training” as a single question.
Sources, exclusions and how to read this chart
The chart uses the same case records and filters as the table below. A row appears only with a dated source, a substantive U.S. ruling and an explicit fair-use result. It compares the copying that court actually assessed. A complaint, a procedural order or an authorship decision cannot supply a fair-use result.
5 records are outside this comparison: 3 from other or unconfirmed jurisdictions; 1 about authorship or registration; 1 procedural or non-merits events; 0 awaiting a dated source or result classification. They remain available in the case table.
Dates belong to the rulings. A weekly edition or later source check does not mean those rulings were issued this week or establish the latest status of each lawsuit. The collection does not support a win rate, legal-risk score or prediction.
The recorded events.
Showing 8 of 8 dated events Newest event first
| Case & parties | Issue & procedure | What happened | Source record |
|---|---|---|---|
ROSS Intelligence, a legal-research company, copied summaries from Thomson Reuters’ Westlaw service for a competing search tool. US · Third Circuit Docket 25-2153 | Training a competing non-generative legal search tool Legal reference text Appeal ruling Appellate merits ruling on interlocutory appeal | Partial summary judgment affirmed The Third Circuit affirmed partial summary judgment for Thomson Reuters: the headnotes at issue were copyrightable and ROSS's copying was not fair use. The opinion distinguishes ROSS's non-generative search engine from generative models. Scope of this eventFair-use context: Rejected for the use at issue An interlocutory appeal, not a declaration that every AI-training use infringes. The PDF says filed Sept. 29; the court's publication list shows Sept. 30. Use the opinion's date. Subsequent docket developments not checked. | Court opinion Read sourceSource checked |
GEMA, which manages songwriters’ rights, challenged the use of songs by Suno, an AI music company. Germany · Munich I Docket 42 O 763/25 | Training / storage / outputs Music compositions District court ruling District court merits ruling | Infringement and related relief ordered Munich I ordered relief concerning six compositions, including US training reproductions, storage in model parameters and public communication in Germany. The judgment addresses Suno versions 3.5 and 4. Scope of this eventFair-use context: Separate jurisdiction; not pooled with US fair-use outcomes A German judgment applying the relevant national laws, not binding US precedent. Does not resolve other music-rightsholders' claims. Subsequent appellate status has not been verified. | Court judgment in Bavaria official legal database Read sourceSource checked |
GEMA, which manages songwriters’ rights, challenged song lyrics reproduced by OpenAI’s ChatGPT. Germany · Munich I Docket 42 O 14139/24 | Memorization / outputs Song lyrics District court ruling District court merits ruling | Core copyright claims substantially upheld Munich I substantially granted GEMA's claims for injunctive relief, information and damages over memorized lyrics and outputs, while rejecting a personality-rights claim concerning misattribution. Scope of this eventFair-use context: German/EU copyright exceptions; not a US fair-use result The court release states that the judgment was not final at publication. Subsequent appellate status has not been verified. | Court press release Read sourceSource checked |
Photo licensing company Getty Images challenged image generator Stability AI over copyright and watermarks. UK · High Court Docket IL-2023-000007 [2025] EWHC 2863 (Ch) | Training, model importation and outputs Images / trademarks Trial judgment Trial judgment on narrowed claims | Limited trademark success; secondary copyright claim failed Getty had withdrawn the primary training and output copyright claims. The court rejected its secondary-infringement claim and found limited, historical trademark infringement involving watermarks. Scope of this eventFair-use context: Not a US fair-use case; primary training claim withdrawn Do not summarize this as a ruling that AI training is lawful. The court expressly did not decide the withdrawn training claim. Later proceedings not checked. | Court judgment Read sourceSource checked |
Authors including Sarah Silverman challenged Meta’s use of their books to train its Llama AI models. US · N.D. California Docket 3:23-cv-03417 | LLM training Books District court ruling District court merits ruling | Summary judgment for Meta The court granted Meta summary judgment on the authors' training claim on the record presented. The decision did not establish that all LLM training is fair use. Scope of this eventFair-use context: Accepted on the plaintiffs' record Historical merits event; subsequent proceedings not checked. Distinguish the evidentiary record from a general legal rule. | U.S. Copyright Office Fair Use Index summary Read sourceSource checked |
Authors challenged how Anthropic, the maker of Claude, obtained and used books to train its AI. US · N.D. California Docket 3:24-cv-05417 | LLM training / library acquisition Books District court ruling District court merits ruling | Training and library uses separated The court treated model training, digitizing lawfully purchased books, and retaining pirated library copies separately. Training and the purchased-book digitization were fair use; the pirated library copies were not. Scope of this eventFair-use context: Mixed by use A historical ruling, not current case status. Do not label the entire lawsuit an unqualified AI victory. | U.S. Copyright Office Fair Use Index summary Read sourceSource checked |
Inventor Stephen Thaler sought copyright registration for an image that named an AI system as its sole author. US · D.C. Circuit Docket 23-5233 | Copyright registration / human authorship AI-generated artwork Appeal ruling Appellate authorship ruling | Registration denial affirmed The D.C. Circuit affirmed denial of the application that named the Creativity Machine as sole author. The court held that the Copyright Act requires eligible works to originate with a human author. Scope of this eventFair-use context: Not at issue Do not treat this as a ruling that human-created work assisted by AI can never be copyrighted. This record preserves the appellate event, not subsequent certiorari history. | Court opinion Read sourceSource checked |
Record labels accused Suno, an AI music company, of copying recordings to train its song generator. US · D. Massachusetts Docket 1:24-cv-11611 | AI music training Sound recordings Complaint filed Complaint filed | Infringement alleged The filed complaint alleges unauthorized copying of sound recordings for Suno's model training. Universal, Sony and Warner-affiliated labels are named plaintiffs. Scope of this eventFair-use context: Not decided by this document Allegations are not findings. Warner later announced settlement of its dispute; this does not establish the posture or resolution of the other labels' claims. Company IDs roll up named label subsidiaries to their groups. | Filed complaint hosted by plaintiffs' trade association Read sourceSource checked |
How to read this collection
One row is one documented event. The date belongs to that event. A later source check does not establish the latest status of the lawsuit. Several records could concern the same dispute.
Procedure and conduct matter. A complaint records allegations. A trial or appeal ruling addresses the issues before that court. Conduct filters describe the record’s subject, including claims that were withdrawn or left undecided.
Jurisdictions remain distinct. German and UK copyright decisions are not pooled into a US fair-use result. This is a selected source collection, not a complete docket search or a prediction of an individual case’s outcome.