DataPacts

Authors v. Apple (Hendrix, Martinez-Conde): Training Data Claims and Status

Checked

At a glance

Authors v. Apple is a consolidated copyright class action in the U.S. District Court for the Northern District of California (Hendrix et al. v.

Independent reference. Every buyer is covered on the same terms, every figure carries its source and check date, and no buyer pays to be listed or placed higher.

Request an independent introduction

Tell us what data your company holds. We reply with the buyers that fit and what their terms leave out. Your details reach a buyer only after you agree to a named introduction.

IndependentYour details stay with us until you agreeNot selling is a valid outcome

Apple Inc., No. 4:25-cv-07558-YGR) in which authors allege that Apple copied their books, including through a pirated-books dataset, to train its OpenELM and Apple Foundation language models, and as of November 17, 2026 the case is in discovery with no ruling on fair use or class certification.

Three suits filed on September 5, October 9 and October 22, 2025 were consolidated on November 14, 2025, and the operative pleading is the First Amended Consolidated Class Action Complaint of November 2, 2026 (Dkt. 130).

Every statement here about Apple's conduct is an allegation, not a finding. This page is part of the independent reference on AI training data licensing, and DataPacts is not affiliated with the court, the parties or counsel, files no claims and gives no legal advice.

Authors v. Apple at a Glance: Court, Dockets, Parties and Status

The case is pending before Judge Yvonne Gonzalez Rogers, with discovery referred to Magistrate Judge Ajay S. Krishnan, and the court has issued no ruling on the merits in the entries read as of November 17, 2026. The table gives the case facts as of 2026-11-17 with the court document behind each.

Field Value Source
Case Hendrix et al. v. Apple Inc. (consolidated with Martinez-Conde v. Apple Inc. and Alexander v. Apple Inc.) Dkt. 27 (Martinez-Conde docket), Dkt. 42 (Hendrix docket)
Court U.S. District Court, N.D. Cal., Oakland Division Dkt. 130 caption
Dockets Hendrix 4:25-cv-07558-YGR (lead; all filings since 2025-11-14); Martinez-Conde 4:25-cv-08695-YGR (filed as 3:25-cv-08695-PHK); Alexander 4:25-cv-09090-YGR Dkt. 27
Judges Gonzalez Rogers, J.; discovery referred to Krishnan, M.J. Dkt. 27, Dkt. 92
Filed Hendrix 2025-09-05; Martinez-Conde 2025-10-09; Alexander 2025-10-22 Dkt. 1; Dkt. 27
Named plaintiffs, operative complaint Grady Hendrix, Jennifer Roberson, John Hornor Jacobs, Eboni McKinnon (Mckinnon Smith in Dkt. 130) Dkt. 130
Defendant Apple Inc. (SEC CIK 0000320193) Dkt. 130; SEC EDGAR
Claim one count, direct copyright infringement, 17 U.S.C. 501; jury demanded Dkt. 81, Dkt. 130
Relief sought statutory damages, compensatory damages, restitution, disgorgement, injunction, destruction of models and copies under 17 U.S.C. 503(b) Dkt. 130 prayer
Status pending; discovery; no ruling on fair use or class certification in the entries read Dkt. 102, 114, 118, 123

What the Authors Allege About OpenELM and Apple Intelligence

The complaints allege that Apple trained its OpenELM models on a copy of a pirated-books dataset called Books3, that it likely trained its Apple Intelligence Foundation Language Models on the same data, and that its web crawler Applebot added more copyrighted text.

The table sets each allegation beside Apple's own published wording as of 2026-11-17; the middle column is Apple's statement, not a finding.

Model or source What the complaint alleges What Apple's own publication says Source and check date
OpenELM (April 2024) trained on a copy of Books3 through the RedPajama Books subset (Dkt. 81 paras. 30 to 47) pre-training dataset contains RefinedWeb, deduplicated PILE, a subset of RedPajama and a subset of Dolma v1.6, about 1.8 trillion tokens Hugging Face apple/OpenELM model card, checked 2026-11-17; Dkt. 81
Foundation Language Models (July 2024 paper) on information and belief trained on the same pirated data (paras. 48 to 61) three sources: data licensed from publishers (a limited amount), curated public or open-source datasets, and the Applebot crawl; licensed data used in later pre-training stages arXiv 2407.21075 section 3.1, checked 2026-11-17; Dkt. 81
Applebot crawl nearly nine years of scraping before the 2024 disclosure; may reach shadow libraries (paras. 3, 62 to 65) Apple respects opt-out by standard robots.txt directives arXiv 2407.21075 section 3.1, checked 2026-11-17; Dkt. 81

OpenELM was released in April 2024, and Apple's model card lists its pre-training data as RefinedWeb, deduplicated PILE, a subset of RedPajama and a subset of Dolma v1.6, about 1.8 trillion tokens (Apple's statement, checked 2026-11-17). The card does not name Books3.

The complaint alleges a chain from that list to Books3. It says the paper behind OpenELM shows the Books subset of RedPajama, that this subset was a copy of Books3 (paras. 34 to 36, 45, 46), and that OpenELM was therefore trained on Books3. It attributes to the dataset's compiler a size of about 196,640 books (para. 38).

The history of the dataset, including its removal, sits on the page about what Books3 is and where it came from, and this page states no removal date.

Apple's Foundation Models paper of July 2024 names three data sources: data licensed from publishers, curated public or open-source datasets, and data crawled by Applebot.

The complaint alleges on information and belief that the open-source category includes Books3 (paras. 48 to 61). Apple's paper describes the licensed data as limited and used in later pre-training stages (checked 2026-11-17).

The plaintiffs seek statutory damages, an injunction and the destruction of Apple's models and copies under 17 U.S.C. 503(b), with a jury trial, on one count of direct infringement under 17 U.S.C. 501 (Dkt. 81, Dkt. 130). These are requests for relief, and the court has granted none.

On Applebot, the complaint alleges nearly nine years of web scraping before Apple said in 2024 that it used the data for training. The crawler page explains how the opt-out works, and this case does not decide it.

Authors v. Apple Docket Entries, 2025 to 2026

Fifteen docket entries carry the case from the complaint of September 5, 2025 to the amended complaint of November 2, 2026. The table lists them as of 2026-11-17, with the court document that holds each. Entries without an archived court PDF were not read, and the docket sheet itself was not available.

Date Dkt. Entry What it did or reported Source
2025-09-05 1 Class action complaint (Hendrix, Roberson) opens the case; OpenELM and Foundation Models claims RECAP archive, Dkt. 1
2025-10-21 37 Related-case order Martinez-Conde related to Hendrix Dkt. 37
2025-11-05 40 Related-case order Alexander related Dkt. 40
2025-11-14 42 Order granting stipulation to consolidate, as amended by the court three actions consolidated; filings on the Hendrix docket; the other two administratively closed Dkt. 27 (Martinez-Conde), Dkt. 42
2026-01-30 79 Order appointing interim co-lead counsel William Dreher and Rohit Nath appointed for one year; terminates Dkt. 57, 59, 60 Dkt. 79
2026-02-13 81 Consolidated Class Action Complaint four named plaintiffs; one count Dkt. 81
2026-03-30 83 Apple's Answer defenses as recited in Dkt. 88 Dkt. 83, Dkt. 88
2026-05-04 88 Joint Case Management Statement and Rule 26(f) report both sides' positions Dkt. 88
2026-05-11 92 Order of reference for discovery Magistrate Judge Krishnan Dkt. 92
2026-06-03 102 Order on timing and staging of summary judgment and class certification summary judgment before any class certification schedule Dkt. 102
2026-06-29 114 Case management and pretrial order schedule through 2027-12-03 Dkt. 114
2026-07-20 118 Order on joint discovery letter brief (Dkt. 117) substantial completion 2026-10-15 Dkt. 118
2026-08-21 123 Order on joint discovery letter brief (Dkt. 119) phase limited to named plaintiffs; Apple to identify all datasets and models with asserted works; hearing 2026-08-25 Dkt. 123
2026-09-08 129 Order denying stipulation without prejudice (Dkt. 128) training-data inspection protocol not entered as an order Dkt. 128, Dkt. 129
2026-11-02 130 First Amended Consolidated Class Action Complaint adds book-scanning and repository allegations Dkt. 130

The court related the three suits in October and November 2025 and consolidated them on November 14, 2025, with all later filings made on the Hendrix docket.

Martinez-Conde was filed as 3:25-cv-08695-PHK and carries a changed header after reassignment. The Alexander caption also names two Apple executives, Craig Federighi and John Giannandrea, as defendants (Dkt. 27), while the consolidated complaints name Apple alone.

The stipulation (Dkt. 27 on the Martinez-Conde docket, Dkt. 42 on the Hendrix docket) was granted as amended, and the other two actions were administratively closed for statistical purposes, not dismissed.

Leadership motions of November 21, 2025 were resolved on January 30, 2026, when the court appointed William Dreher and Rohit Nath interim co-lead counsel for one year (Dkt. 79). The four named plaintiffs of the consolidated complaints do not include Martinez-Conde, Macknik or Alexander, and the entries read do not say why.

The Consolidated Complaint of February 2026 and the Amended Complaint of November 2026

The Consolidated Class Action Complaint (Dkt. 81, filed 2026-02-13) named four authors as plaintiffs and pleaded one count, direct copyright infringement, against Apple alone, and Apple answered on 2026-03-30 (Dkt. 83). The plaintiffs are Grady Hendrix, Jennifer Roberson, John Hornor Jacobs and Eboni McKinnon (spelled Mckinnon Smith in Dkt. 130), each pleading registered works.

Dkt. 130 (November 2, 2026, the last day to amend under Dkt. 114) keeps the claims and adds allegations that Apple obtained books by scanning them, made further copies in other formats and stored them in central repositories (paras. 4, 88 to 95).

Figures in paragraphs 88 and 90 are redacted, and Apple asked on November 9, 2026 (Dkt. 133, answering Dkt. 132) to keep them sealed, so this page states no number from them. Apple's defenses appear here only as Dkt. 88 recites them.

Discovery Orders From May to September 2026

Discovery was referred to Magistrate Judge Ajay S. Krishnan on May 11, 2026 (Dkt. 92), and his orders so far set the scope of the current phase, a document deadline and a limit on converting a party agreement into an order.

The first order, Dkt. 118 (2026-07-20), adopted in large part the plaintiffs' proposal: for requests served by 2026-05-29, substantial production by 2026-10-15, with exceptions for matters beyond a producer's control and rolling production expected.

Dkt. 123 (2026-08-21, on Dkt. 119) limited the current phase to the named plaintiffs' claims and excluded class discovery. It also ordered Apple to identify all datasets and associated language models containing the asserted works, not only those named in the complaint. A Zoom hearing was set for 2026-08-25, and its outcome was not read.

Dkt. 129 (2026-09-08) denied without prejudice a stipulation on a training-data inspection protocol (Dkt. 128), saying the court would generally enforce the parties' agreement. The protocol's content is not described in the order.

Apple's Stated Position: Fair Use and Individual Issues

In its section of the Joint Case Management Statement of May 4, 2026 (Dkt. 88), Apple states that its training of large language models was fair use and that the defense is dispositive of the entire case.

Apple cites two earlier decisions of the same district, Bartz at 787 F. Supp. 3d 1007 and Kadrey at 788 F. Supp. 3d 1026, and states that OpenELM was never commercially released and was made for research, while its Foundation Models are integrated into Apple products.

The doctrine and the other rulings are covered on the page on the four factors and the other rulings.

Apple also states that the putative class raises individualized issues: ownership, registration, publisher licenses, market harm and damages. The defenses Dkt. 88 recites include express or implied license, waiver, laches, estoppel and the statute of limitations.

The plaintiffs state in the same filing that Apple used pirated datasets and that its models compete with their works. The court has decided none of it, and this page does not suggest that either side is likely to prevail.

What the Court Has Scheduled and What It Has Left Open

The court has set the end of merits fact discovery for April 30, 2027, a summary judgment and Daubert hearing for December 3, 2027, and no date for class certification or trial. The table lists the events as of 2026-11-17 from the case management order (Dkt. 114) and the staging order (Dkt. 102).

Event Date Status Source
Last day to amend pleadings or join parties 2026-11-02 passed; Dkt. 130 filed that day Dkt. 114
Substantial completion of documents for requests served by 2026-05-29 2026-10-15 passed Dkt. 118
Close of merits fact discovery (individual plaintiffs' claims, Apple's defenses) 2027-04-30 upcoming Dkt. 114
Expert reports: opening, opposition, rebuttal 2027-05-28, 2027-06-25, 2027-07-19 upcoming Dkt. 114
Case management conference 2027-07-26, 2:00 p.m. upcoming Dkt. 114
Expert discovery cutoff 2027-07-30 upcoming Dkt. 114
Dispositive and Daubert motions: opening, opposition, reply 2027-08-27, 2027-09-24, 2027-10-29 upcoming Dkt. 114
Hearing on dispositive and Daubert motions 2027-12-03 upcoming Dkt. 114
Class certification no date; after summary judgment, if needed open Dkt. 102
Trial no date in the entries read open Dkt. 114

Dkt. 102 (2026-06-03) ordered summary judgment resolved before any class certification schedule, if one is needed. Dkt. 85 (2026-04-17) shows that Apple expressly waived the one-way intervention rule for that sequence, while the plaintiffs had proposed class certification dates in 2028.

Dkt. 114 (2026-06-29) sets the other dates. The deadlines of October 15 and November 2, 2026 are shown as passed, and the page does not state whether any party complied.

What the Apple Case Signals to Holders of Licensable Data

The case records a claim, not a market: the authors allege Apple did not license their books, while Apple's own paper says it licensed a limited amount of data from publishers.

That paper (checked 2026-11-17) places licensed data in the continued and context-lengthening stages of pre-training, and the complaint pleads a licensing market for AI training data and no payment to the plaintiffs (Dkt. 81 para. 6), which is an allegation. Terms and prices for books are covered under book license fees and disputes.

The pending allegations of acquisition from shadow libraries are provenance questions for any seller or buyer of book data. No outcome exists yet, and nothing here calls for action by a holder of data.

Nothing in this case is a price or a license rate. No license term or Apple licensing value comes from the record read, so the page states none.

Four other book-training cases in the tracker share the Northern District of California or the same pirated-books datasets, and each has its own page. The tracker lists all of them under every tracked case with docket and status, and the table gives the four closest to this one as of 2026-11-17.

Case Court and docket Status Page
Bartz v. Anthropic N.D. Cal. 3:24-cv-05417-AMO settled; final approval and judgment 2026-07-20 (Dkt. 680) Bartz v. Anthropic
Kadrey v. Meta N.D. Cal. 3:23-cv-03417-VC training claim decided for Meta on its record 2025-06-25 (Dkt. 598); first coordinated trial set for 2027-05-03 Kadrey v. Meta
In re Salesforce (Tanzer v. Salesforce) N.D. Cal. 3:25-cv-08862-CRB pending; XGen models allegedly trained on The Pile, RedPajama and Books3; class certification motion due 2027-12-03 In re Salesforce
Huckabee v. Bloomberg S.D.N.Y. 1:23-cv-09152-MMG motion to dismiss denied 2025-11-24 (Dkt. 98); fact discovery closes 2027-05-27 Huckabee v. Bloomberg

Related pages:

More About the Apple Authors Case

Pending Versus Decided: What Is Open as of November 17, 2026

As of November 17, 2026 the court has decided procedural matters only: relation, consolidation, leadership, the order of motions, discovery scope and deadlines. The highest entry read is Dkt. 133 (2026-11-09), and entries without an archived PDF were not read.

  • Decided: relation of the three suits, consolidation, interim co-lead counsel, summary judgment before class certification, discovery scope and the document deadline.
  • Open: fair use, class certification, the merits, Apple's response to Dkt. 130 (not in the entries read), the outcome of the 2026-08-25 hearing and trial.

This page is information, not legal advice.

Did authors sue Apple over AI training?

Yes: Hendrix and Roberson filed on 2025-09-05, two more suits followed, and all were consolidated on 2025-11-14. The consolidated case is Hendrix et al. v. Apple Inc., No. 4:25-cv-07558-YGR.

Is there a settlement with Apple?

No settlement appears in the entries read as of 2026-11-17. Settlements in other Apple matters are not part of this case.

Has the court decided whether Apple's training was fair use?

No: summary judgment is to be decided before class certification, with the hearing set for 2027-12-03. Apple states that its training was fair use (Dkt. 88), and the court has not ruled.

Is Apple Intelligence trained on pirated books?

That is an allegation made on information and belief, and no court has found liability. Apple states its position in Dkt. 88, and its paper describes its training data in three categories.

Can I join the case or file a claim?

No claim process exists and no class is certified. The court has not approved any notice or settlement. Authors weighing a license can instead read about licensing books for AI training.

The figures and dates on this page come from court documents and Apple's own publications, and how figures are sourced and dated is set out on the methodology page.