As of November 16, 2026 both sides have moved for summary judgment on fair use, the Class Plaintiffs' reply is due on November 20, 2026, and no ruling, class certification or settlement is recorded. The pleadings were settled by two orders of October 27, 2025.
The News Plaintiffs, led by The New York Times, are a separate track in the same MDL and have their own page. DataPacts is not affiliated with any party and gives no legal advice; this page belongs to the independent reference on AI training data licensing.
Dockets, dates and quotations come from the court filings named in each table, read on November 16, 2026, and how figures are sourced and dated is set out on the methodology page.
The Authors' Class Actions at a Glance: Courts, Dockets, Parties and Status
No judgment exists in the authors' case: the pleadings were settled by two orders of October 27, 2025, and both sides have asked the court to decide fair use on summary judgment. The table lists the case facts as of 2026-11-16.
| Field | Value | Source |
|---|---|---|
| Case | Authors Guild et al. v. OpenAI Inc. et al. and related class actions; consolidated complaint captioned David Baldacci et al. v. OpenAI Inc. et al. | KB; Dkt. 183 |
| Court | U.S. District Court, S.D.N.Y., within MDL No. 3143 | JPML Dkt. 85 |
| Dockets | Authors Guild 1:23-cv-08292-SHS-OTW; Alter 1:23-cv-10211-SHS-OTW; Tremblay N.D. Cal. 3:23-cv-03223; Silverman N.D. Cal. 3:23-cv-03416; Chabon N.D. Cal. 3:23-cv-04625 (S.D.N.Y. 25-cv-3291); MDL 1:25-md-03143-SHS-OTW | JPML Dkt. 85; Dkt. 707 |
| Judges | Sidney H. Stein, J.; Magistrate Judge Ona T. Wang (before transfer: Araceli Martínez-Olguín, J., N.D. Cal.) | Dkt. 701; Dkt. 709; N.D. Cal. Dkt. 104 |
| Plaintiffs | 22, including The Authors Guild and 14 individual authors; putative class | Dkt. 183 |
| Defendants | ten OpenAI entities and Microsoft Corporation | Dkt. 183 |
| Claims | direct, vicarious and contributory copyright infringement (training copies, downloads, outputs) | Dkt. 183; Dkt. 707 |
| In-scope models | GPT-3, GPT-3.5, GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, GPT-4o, GPT-4o mini | Dkt. 707 |
| Class certification | none found; proposed class only | Dkt. 183; Dkt. 701 |
| Status | cross-motions for summary judgment briefed; Class Plaintiffs' reply due 2026-11-20; no ruling | Dkt. 2312; Dkt. 2495 |
| Settlement | none recorded | KB case record; MDL docket read to Dkt. 2505 |
From Five Author Suits to One Consolidated Class Complaint
Five author suits came first: Authors Guild and Alter in the Southern District of New York, and Tremblay, Silverman and Chabon in the Northern District of California. Each was a putative class action by authors against OpenAI entities, and Microsoft was a defendant in the New York suits.
The Judicial Panel on Multidistrict Litigation centralized all five with seven other actions on April 3, 2025 (Dkt. 85), and Chabon now carries S.D.N.Y. No. 25-cv-3291 (Dkt. 707, n. 2). Membership of the MDL and the News Plaintiffs are set out on the page for In re OpenAI copyright infringement litigation and its member cases.
On May 22, 2025 the court ordered one consolidated complaint limited to the products and claims already pleaded. The Class Plaintiffs filed it on June 13, 2025 (Dkt. 183), captioned David Baldacci et al. v.
OpenAI Inc. et al., with 22 plaintiffs including The Authors Guild and 14 individual authors, and 11 defendants: ten OpenAI entities and Microsoft.
The 2024 California Ruling in Tremblay and Silverman
Before the transfer, Judge Araceli Martínez-Olguín of the Northern District of California dismissed most of the California authors' claims on February 12, 2024, with leave to amend, and let the direct infringement claim and the unfair-prong UCL claim go forward (Dkt. 104).
The dismissed claims were vicarious infringement (no output was alleged to be substantially similar to a book), removal of copyright management information under DMCA section 1202(b), the unlawful and fraudulent prongs of the UCL, negligence and unjust enrichment.
Direct infringement was not challenged. The First Consolidated Amended Complaint of March 13, 2024, "In re OpenAI ChatGPT Litigation" (master file 23-cv-3223-AMO, Dkt. 120), pleaded two counts: direct infringement and UCL unfair competition. No later California ruling is stated here.
What the Class Plaintiffs Claim: Three Counts Over Books
The Consolidated Class Action Complaint pleads three counts: direct copyright infringement against OpenAI and Microsoft, vicarious infringement and contributory infringement. The table gives each count as pleaded on 2025-06-13, with its status on 2026-11-16.
| Count | Claim | Against | Rests on | Status 2026-11-16 | Source |
|---|---|---|---|---|---|
| I | Direct infringement, 17 U.S.C. 501 | OpenAI and Microsoft | downloads, training copies, outputs | output part survived Dkt. 701; summary judgment pending | Dkt. 183 paras. 308 to 327; Dkt. 701 |
| II | Vicarious infringement | Microsoft, OpenAI Inc., OpenAI GP LLC | control and financial benefit | pending | Dkt. 183 |
| III | Contributory infringement | Microsoft and nine OpenAI entities | investment, supercomputing system, control | pending | Dkt. 183; Dkt. 707 |
The court confirmed that the case covers books only and OpenAI-trained models only (Dkt. 707). Every statement in this section and the two below describes what the authors allege, not a finding of any court.
The Copying and Training Allegations
The authors allege that OpenAI and Microsoft reproduced their books from shadow libraries and used the copies to build datasets for training GPT models, without permission or payment.
The complaint alleges that OpenAI has admitted the Books1 and Books2 datasets were downloaded from LibGen, and it names LibGen, Z-Library and Bibliotik as sources (Dkt. 183, paras. 112 to 119 and 311).
The complaint also points to Books3, and what that dataset contains is explained on the page about Books3. OpenAI's position that no in-scope model was trained on Books3 appears below, in the summary-judgment section.
The Output Claim and the Relief the Authors Seek
Besides the training copies, the authors claim that some ChatGPT outputs infringe their books directly, a claim the court let proceed in October 2025. The complaint of June 13, 2025 asks for the following relief (Dkt. 183, para. 328):
- class certification and appointment of class counsel;
- an injunction, including against operating models trained on class works without authorization;
- actual damages and profits;
- statutory damages of up to USD 150,000 per infringed work, at the authors' election, as pleaded on 2025-06-13;
- disgorgement, fees and interest.
These are requests in a pleading, not an expected award.
The Proposed Class and Where Class Certification Stands
The authors sue for a proposed class of book copyright owners, and no order certifying that class was found in the court record read on November 16, 2026.
The class is defined in the complaint under Rules 23(b)(3), 23(b)(2) and 23(c)(4) (Dkt. 183, para. 294) as all legal or beneficial owners of copyrighted works that meet four conditions:
- the work was registered with the U.S. Copyright Office within five years of first publication;
- the work was downloaded or otherwise reproduced by OpenAI or Microsoft;
- the work was registered before that copying, or within three months of first publication;
- the work is assigned an ISBN or an ASIN.
The complaint alleges that the class consists of more than one hundred thousand authors and copyright holders (para. 296), which is an allegation and not a finding. The October 2025 opinion still calls it "a proposed class" (Dkt. 701), and no certification motion or schedule was located.
Authors v. OpenAI Docket Entries, 2024 to 2026
Eleven filings carry the authors' case from the California ruling of 2024 to the summary-judgment briefing of November 2026. The table lists them as of 2026-11-16 with the document each rests on.
| Date | Court and entry | What it decided or shows | Source |
|---|---|---|---|
| 2024-02-12 | N.D. Cal. Dkt. 104 | motions to dismiss granted in part: vicarious, DMCA, UCL unlawful and fraudulent, negligence, unjust enrichment dismissed with leave to amend; unfair-prong UCL proceeds | RECAP archive PDF |
| 2024-03-13 | N.D. Cal. Dkt. 120 | First Consolidated Amended Complaint: direct infringement and UCL | RECAP archive PDF |
| 2025-04-03 | JPML Dkt. 85 | transfer order: five author suits among 12 centralized actions | jpml.uscourts.gov |
| 2025-06-13 | MDL Dkt. 183 | Consolidated Class Action Complaint, three counts, proposed class | RECAP archive PDF |
| 2025-10-27 | MDL Dkt. 701 | motion to dismiss output claim denied; no fair-use conclusion | RECAP archive PDF |
| 2025-10-27 | MDL Dkt. 707 | motions to strike granted in part: models limited to seven; copying claim kept | RECAP archive PDF |
| 2025-10-28 | MDL Dkt. 709 | additional discovery on the "LibGen Test File" granted | RECAP archive PDF |
| 2026-10-06 | MDL Dkt. 2090 | expert supplemental report struck; market-dilution paper precluded | CourtListener PDF |
| 2026-10-29 | MDL Dkt. 2312 | public redacted version of OpenAI's opposition to Class Plaintiffs' motion for summary judgment | RECAP archive PDF |
| 2026-11-05 | MDL Dkt. 2494 | defendants oppose more reply words | RECAP archive PDF |
| 2026-11-06 | MDL Dkt. 2496 | order on fifty-five amicus motions | RECAP archive PDF |
No summary-judgment ruling was in the public record on 2026-11-16, and the latest MDL entry read is Dkt. 2505, filed 2026-11-10.
The October 27, 2025 Rulings on Outputs, Downloads and Models
Judge Stein decided two motions on October 27, 2025: he denied OpenAI's motion to dismiss the output-based infringement claim (Dkt. 701) and struck only the allegations about models outside the agreed list (Dkt. 707).
In Dkt. 701 the court held that the complaint states a prima facie claim as to at least some outputs, and that a reasonable jury could find an output summarizing a sequel to a George R.R. Martin novel substantially similar in setting, plot and characters. The opinion "makes no conclusions regarding fair use".
In Dkt. 707 the court limited the case to GPT-3, GPT-3.5, GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, GPT-4o and GPT-4o mini, the models taken from Magistrate Judge Wang's discovery ruling in the Authors Guild action. Allegations on GPT-4V, GPT-4.5, GPT-5 and "derivatives" and "successors" of any model were struck.
OpenAI's request to strike what it called a new claim about copies downloaded from shadow libraries was denied because the earlier complaints already alleged the downloading and reproduction. Microsoft's requests concerning GPT-1, GPT-2, GPT-4o and GPT-4o mini were denied, with discovery into its products remaining subject to proportionality.
Discovery and the Summary-Judgment Briefing on Fair Use
Both sides have moved for summary judgment: the Class Plaintiffs ask the court to reject fair use (MDL ECF 1848), and OpenAI asks it to find its use fair as a matter of law (MDL ECF 1718).
Discovery entries tied to this track include the order of October 28, 2025, in which Magistrate Judge Wang granted the Class Plaintiffs' request for more discovery on the "LibGen Test File" (Dkt. 709), and the order of October 6, 2026 precluding the Class Plaintiffs from relying on their expert's market-dilution paper (Dkt. 2090).
In the public redacted version of its opposition, filed October 29, 2026 (Dkt. 2312), OpenAI states the following, and each point is OpenAI's own statement:
- only GPT-3 and GPT-3.5 among the in-scope models were trained on material downloaded from LibGen, and OpenAI stopped using LibGen books in late 2021;
- no in-scope model was trained on Books3;
- its expert found 14 instances of verbatim or near-verbatim regurgitation in about 95 million outputs from 20 million conversations;
- the distribution claim was never pleaded in the complaint;
- Bartz and Kadrey rejected lost-license harm.
The four factors and other courts' rulings are explained on the page about fair use and AI training in US law, and this page does not weigh them.
The Class Plaintiffs' reply is due on November 20, 2026 with a 7,500-word limit (Dkt. 1664 and 2049, as their letter Dkt. 2495 states), and the defendants opposed their request for 2,500 more words (Dkt. 2494).
On November 6, 2026 the court ruled on fifty-five motions for leave to file amicus briefs (Dkt. 2496). No ruling on summary judgment was recorded as of 2026-11-16.
What the Authors' Case Shows Book Rights Holders
The authors' case shows that the dispute over book training turns on how the copies were obtained and which models used them, and that no court has yet priced that use.
The record separates the copying, training and output claims, and the class definition turns on registration timing and an ISBN or ASIN, so registration records are what that definition relies on.
The Authors Guild, a plaintiff here, holds that standard trade contracts grant no AI training rights, which is its own advocacy position. How authors and publishers license books is covered on the page about authors and publishers licensing books for AI training.
A pending case is not a reason to license or to wait, and not licensing at all remains an option. This page is information, not legal advice.
Where the Authors' Suits Sit Among Other Book-Training Cases
The authors' suits against OpenAI are one of three groups of book-training cases in the tracker, next to the settled Anthropic case and the Meta case decided for Meta on training.
The case tracker of AI copyright lawsuits over training data lists every tracked case with its docket and status, and the table gives the three as of 2026-11-16.
| Case | Court and docket | Status | Source |
|---|---|---|---|
| The New York Times v. Microsoft and OpenAI | S.D.N.Y. 1:23-cv-11195-SHS-OTW (MDL 3143, News Plaintiffs) | summary-judgment stage; no ruling located | JPML Dkt. 85; KB case record |
| Bartz v. Anthropic | N.D. Cal. 3:24-cv-05417-AMO | settled; USD 1.5 billion; final approval 2026-07-20 (Dkt. 680) | RECAP archive PDF |
| Kadrey v. Meta | N.D. Cal. 3:23-cv-03417-VC | training claim decided for Meta on the record 2025-06-25 | CourtListener |
OpenAI's licensing deals belong on its profile as a licensee, covered on the page about OpenAI's licensing deals and data programs.
More About the Authors' Class Actions Against OpenAI
Decided Versus Still Open in the Authors' Case
The pleadings are decided; fair use, class certification and any remedy are open.
- Decided: the California dismissals of February 12, 2024, with leave to amend; the consolidation and the complaint of June 2025; the survival of the output claim (Dkt. 701); the in-scope models (Dkt. 707).
- Open: summary judgment on fair use; the distribution theory; class certification; any relief.
Frequently Asked Questions About Authors Guild v. OpenAI
Can authors sue OpenAI over training on their books?
Yes: the Class Plaintiffs' training and output claims survived the pleadings, and whether the use was fair is briefed, not decided.
Has the court certified a class of authors?
No certification order was found as of 2026-11-16. The complaint seeks one.
Is there a settlement or payout for authors?
None is recorded and no claim process exists. The settled book case in the tracker is Bartz v. Anthropic.
Has the court ruled on fair use in the authors' case?
No: Dkt. 701 reserved the question and the cross-motions await a ruling.
Is this the same case as Authors Guild v. Google?
No: that earlier case concerned Google Books. OpenAI's opposition cites its appellate decision as precedent (Dkt. 2312).
What happened to Tremblay v. OpenAI?
It was narrowed in 2024, consolidated with Silverman and Chabon, and transferred into MDL 3143 on 2025-04-03. Its current S.D.N.Y. number is not stated.