DataPacts

Huckabee v. Bloomberg (BloombergGPT and Books3): Claims, Rulings and Status

Checked

At a glance

Huckabee v. Bloomberg is a proposed class action that Mike Huckabee and other authors filed in the U.S. District Court for the Southern District of New York on October 17, 2023 (No. 1:23-cv-09152-MMG), claiming that Bloomberg L.P. and Bloomberg Finance L.P. infringed their copyrights by training the BloombergGPT model on Books3. On November 24, 2025, Judge Margaret M. Garnett denied Bloomberg's motion to dismiss in full and held that fair use could not be decided on the complaint without a factual record.

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

As of November 17, 2026, the case is in fact discovery, which now closes on May 27, 2027 (Dkt. 139).

The claim in the amended complaint is one count of direct copyright infringement, and no court has found infringement or ruled on fair use. DataPacts is not affiliated with any party and gives no legal advice; this page belongs to the independent reference on AI training data licensing.

Dates, docket numbers and quotations come from the court filings named in each table, read on November 17, 2026; how figures are sourced and dated is set out on the methodology page.

Huckabee v. Bloomberg at a Glance: Court, Docket, Parties and Status

No ruling on the merits exists: the only decision so far is the 2025-11-24 denial of Bloomberg's motion to dismiss, and the parties are exchanging discovery under a schedule extended on 2026-11-05. The table lists the case facts as of 2026-11-17.

Field Value Source
Case Huckabee v. Bloomberg L.P. (filed as Huckabee v. Meta Platforms, Inc.) Dkt. 98; Dkt. 1, 96 captions
Court U.S. District Court for the Southern District of New York Dkt. 98
Docket 1:23-cv-09152-MMG Dkt. 98 header
Filed 2023-10-17 Dkt. 1 stamp
Judge Margaret M. Garnett Dkt. 85, 98, 119, 139
Plaintiffs Mike Huckabee, Relevate Group and four other named authors, for a proposed class Dkt. 74 caption
Defendants Bloomberg L.P. and Bloomberg Finance L.P. (originally also Meta, Microsoft and The EleutherAI Institute) Dkt. 74; Dkt. 1
Claim one count of direct copyright infringement, 17 U.S.C. 106 Dkt. 74; Dkt. 98
Model and dataset BloombergGPT; Books3 (part of The Pile) Dkt. 98
Key ruling motion to dismiss denied in full, 2025-11-24 Dkt. 98
Status fact discovery; closes 2027-05-27; summary judgment conference 2027-08-24; no class certified; no trial date Dkt. 139

From Five Defendants to One: How the Complaint Narrowed

The case began against five defendants, Meta Platforms, Bloomberg L.P., Bloomberg Finance L.P., Microsoft Corporation and The EleutherAI Institute, under the caption Huckabee v.

Meta Platforms, Inc., and with claims for direct and indirect copyright infringement, DMCA violations, conversion, negligence and unjust enrichment. The complaint of 2023-10-17 (Dkt. 1) is described that way in Bloomberg's motion papers (Dkt. 87, p. 2).

On 2023-12-15 counsel for the EleutherAI Institute reported that the plaintiffs agreed in principle to dismiss it without prejudice (Dkt. 59), and Microsoft reported their agreement to sever and transfer the claims against Meta and Microsoft to the Northern District of California (Dkt. 60).

The opinion summarizes the result as dismissing one defendant and severing claims against two (Dkt. 98).

After Bloomberg's pre-motion letter the court ordered an amended complaint, and the amended complaint of 2024-01-24 (Dkt. 74) keeps one count of direct infringement against the two Bloomberg entities. No finding about EleutherAI exists, and the severed claims after transfer are not covered here.

What the Authors Claim About BloombergGPT

The authors claim that their books were among the copyrighted works in Books3, a component of the larger dataset called The Pile, and that Bloomberg copied and used Books3 to train BloombergGPT without permission or payment. Everything here is the plaintiffs' allegation, never a finding.

The origin of the dataset, its counts and its removal from hosting are set out on the page on Books3, the book dataset at issue.

The Training Allegations and the Spokesperson Email

The amended complaint alleges that Bloomberg announced BloombergGPT on March 30, 2023 as a 50-billion-parameter model "built from scratch for finance," trained on a corpus of over 700 billion tokens that included Books3. The Books3 allegation cites a review of Bloomberg's research paper (para. 48, n.17), which the court read as a paper Bloomberg released stating that it used Books3 (Dkt. 98, pp. 2 to 3).

The complaint adds that Books3 let Bloomberg build the model "faster and more efficiently" without paying for source material (para. 49).

The complaint also alleges that a Bloomberg spokesperson confirmed by email that Books3 was used to train the initial model and said it would not be used for future versions (para. 50), and that the works are "baked into" later versions (paras. 51 to 52). Bloomberg's answer says the email speaks for itself and denies the remaining allegations.

The Proposed Class and the Relief Sought

The authors sue for a class of all persons or entities in the United States that own a U.S. copyright in any work used as training data for Bloomberg's model from October 2020 to the present, which they estimate at "at least tens of thousands" of members (paras. 61, 64). The amended complaint asks for:

  • class certification;
  • declaratory relief and an injunction against the alleged practices;
  • statutory and other damages under 17 U.S.C. 504;
  • punitive damages, statutory and civil penalties and interest;
  • costs and attorneys' fees, with a jury demand.

No amount is pleaded, and no class has been certified as of 2026-11-17.

The November 2025 Ruling: Motion to Dismiss Denied in Full

Judge Margaret M. Garnett denied Bloomberg's motion to dismiss in full on November 24, 2025 (Opinion and Order, Dkt. 98), rejecting both of Bloomberg's arguments: that the complaint did not identify which works were in Books3, and that fair use defeated the claim on its face.

The motion was filed on 2024-03-22 (Dkt. 86, 87), opposed on 2024-04-19 (Dkt. 90) and replied to on 2024-05-03 (Dkt. 91), with discovery stayed meanwhile (Dkt. 85).

Bloomberg argued that BloombergGPT was a research project, that the complaint did not allege a public or commercial release or that the model displaced the plaintiffs' books, and that dismissal should be with prejudice (Dkt. 87, pp. 1 to 3).

A denial decides only that the claim may go forward on the allegations; it is not a finding of infringement and not a rejection of fair use.

Why the Pleading Gave Fair Notice

The court held that Rule 8 requires fair notice of the claim, not every factual detail, and that the amended complaint gave it by alleging that Books3 contains the plaintiffs' works, citing Bloomberg's own paper on using Books3 to train BloombergGPT, and stating that the use was unauthorized.

The court rejected Bloomberg's argument that the complaint must say by what acts and when the copying occurred.

On the unlisted works, the court found that the "clear and logical inference" from the allegations is that one or more of each named plaintiff's works were in Books3, which suffices at this stage (Dkt. 98, pp. 4 to 5).

Why the Court Would Not Decide Fair Use on the Complaint

Fair use is an affirmative defense, and the court held that Bloomberg had not shown it was "clearly established on the face" of the complaint, the Second Circuit standard in TCA Television Corp. v. McCollum (839 F.3d 168, 178).

Using the fourth factor as its example, the court listed questions that need a record: whether widespread copying of literary works to train models will flood the market with AI-produced works, whether a derivative market for literary works used in training exists, and whether unlicensed copying harms authors' ability to profit from it.

Footnote 3 notes conflicting Second Circuit precedent on counting a potential licensing market (TCA versus Authors Guild v. HathiTrust) and says the result is the same either way.

On the first factor, footnote 4 says that, drawing inferences for the plaintiffs, the complaint suggests a commercial purpose, since Bloomberg is a finance news company and the model was built "for finance."

That is a pleading-stage inference. How US courts apply the four factors to training is covered on the page on fair use and AI training in US law.

Bloomberg's Answer and Affirmative Defenses

Bloomberg answered the amended complaint on January 27, 2026 (Dkt. 125), admitting that it published a press release and a research paper about "a research LLM called BloombergGPT" on March 30, 2023, and denying the allegations that it trained the model on Books3 and saved money by doing so.

It denies paragraphs 48 and 49, says the spokesperson email "speaks for itself" (paragraph 50) and demands a jury.

The answer pleads 13 affirmative defenses. Five of them, named as Bloomberg pleads them:

  • ownership of exclusive rights (first);
  • invalid or insufficient registrations under 17 U.S.C. 411 to 412 (second);
  • fair use (third);
  • express or implied license (fourth);
  • de minimis copying (tenth).

A defense pleaded is a position, not a ruling.

Huckabee v. Bloomberg Docket Entries, 2023 to 2026

Twelve filings and orders carry the case from the five-defendant complaint to the extended discovery schedule. The table lists them as of 2026-11-17, from court PDFs in the public RECAP archive; entries not in the archive were not read.

Date Entry What it decided or reported Source
2023-10-17 Dkt. 1 complaint five defendants; direct and indirect infringement, DMCA, conversion, negligence, unjust enrichment Dkt. 1; Dkt. 87 p. 2
2023-12-15 Dkt. 59, Dkt. 60 letters plaintiffs agree to dismiss EleutherAI Institute without prejudice and to sever and transfer claims against Meta and Microsoft to N.D. Cal. Dkt. 59; Dkt. 60
2024-01-24 Dkt. 74 amended complaint one count of direct infringement against Bloomberg L.P. and Bloomberg Finance L.P. Dkt. 74; Dkt. 98
2024-03-20 Dkt. 85 order briefing schedule; discovery stayed during briefing Dkt. 85
2024-03-22 to 2024-05-03 Dkt. 86, 87, 90, 91 motion to dismiss, opposition, reply Dkt. 85, 98
2024-08-22 Dkt. 96 Bloomberg's notice of supplemental authority (Santos v. Kimmel) Dkt. 96
2025-05-30 Dkt. 97 plaintiffs' notice of supplemental authority (Copyright Office Part 3) Dkt. 97
2025-11-24 Dkt. 98 Opinion and Order motion to dismiss denied in full Dkt. 98
2026-01-13 Dkt. 119 order discovery stay lifted; schedule adopted Dkt. 119
2026-01-27 Dkt. 125 answer denials; 13 affirmative defenses Dkt. 125
2026-02-20 and 2026-03-23 Dkt. 129, Dkt. 136 protective order; order on production of electronically stored information Dkt. 129; Dkt. 136
2026-11-05 Dkt. 139 joint letter-motion of 2026-11-02 granted; deadlines extended 90 days Dkt. 139

The Discovery Schedule After the November 2026 Extension

On November 5, 2026 the court granted the parties' joint request of November 2 and extended every pending deadline by 90 days, so fact discovery now closes on May 27, 2027 and the summary judgment conference is set for August 24, 2027 (Dkt. 139).

The table compares the January 2026 dates with those after the extension, as of 2026-11-05.

Event Date set 2026-01 Date after 2026-11-05 Source
substantial completion of document production 2026-11-30 2027-03-01 Dkt. 139
requests for admission served 2026-12-15 2027-03-15 Dkt. 139
privilege logs exchanged 2026-12-30 2027-03-30 Dkt. 139
close of fact discovery 2027-02-26 2027-05-27 Dkt. 119; Dkt. 139
opening expert reports 2027-03-03 2027-06-01 Dkt. 120; Dkt. 139
rebuttal expert reports 2027-03-24 2027-06-22 Dkt. 120; Dkt. 139
replies to rebuttal reports 2027-04-07 2027-07-06 Dkt. 120; Dkt. 139
close of expert discovery 2027-04-16 2027-07-15 Dkt. 119; Dkt. 139
summary judgment conference 2027-05-25 2027-08-24, 9:30 a.m. Dkt. 119; Dkt. 139

Merits schedule only: class certification expert discovery and motions follow a summary judgment decision, if necessary (Dkt. 119).

The order of January 13, 2026 (Dkt. 119) lifted the discovery stay and adopted the original dates. The court wrote that the parties "should not expect further extensions" barring unusual circumstances (Dkt. 139).

Bloomberg reserved the right to seek changes if the scope of the infringement claim is unclear at the close of fact discovery (Dkt. 120).

What the Case Shows Holders of Book Rights

The case shows that a court in the Southern District of New York will not resolve fair use for book training on the pleadings, and that the market questions it listed, including whether a market for licensing books for training exists, are left for evidence. The ruling does not say such a market exists.

The parties' supplemental authorities show the two sides of the record. Bloomberg cited Santos v. Kimmel (S.D.N.Y. 2024) for deciding fair use on a motion to dismiss (Dkt. 96, 2024-08-22), and the plaintiffs cited the Copyright Office's pre-publication Part 3 report of 2025-05-09 for treating fair use as fact-intensive (Dkt. 97, 2025-05-30).

That report says knowing use of pirated or illegally accessed works weighs against fair use without being determinative (p. 52); it is covered on the page on the US Copyright Office report on generative AI training.

No outcome is predicted. The case gives authors no payment route today, and nothing here is advice to sue, join or license.

Huckabee v. Bloomberg is the only book-training case in the tracker against a financial-information company, next to the authors' cases against Anthropic, Meta, NVIDIA and Salesforce that also name Books3. Every tracked case is listed in the case tracker of AI copyright lawsuits over training data with docket and status.

Case Court and docket Books3 link Status Page
Bartz v. Anthropic N.D. Cal. 3:24-cv-05417-AMO Books3 among downloaded library copies settled; final approval 2026-07-20 (Dkt. 680), USD 1.5 billion fund; no distribution confirmed (checked 2026-11-17) Bartz v. Anthropic
Kadrey v. Meta N.D. Cal. 3:23-cv-03417-VC Books3 in Llama training per the court training claim decided for Meta 2025-06-25; first trial in related cases 2027-05-03 (Dkt. 879) Kadrey v. Meta
Nazemian v. NVIDIA N.D. Cal. 4:24-cv-01454-JST Books3 alleged for NeMo Megatron discovery; fact discovery closes 2027-02-12 (as of 2026-11-17) Nazemian v. NVIDIA
In re Salesforce (Tanzer v. Salesforce) N.D. Cal. 3:25-cv-08862-CRB Books3 and RedPajama alleged for XGen pending In re Salesforce (Tanzer v. Salesforce)

The table compares four related book-training cases with this one as of 2026-11-17; two further pages give the wider context:

More About Huckabee v. Bloomberg

Information on this page is not legal advice.

Decided Versus Still Open in Huckabee v. Bloomberg

Only the pleading questions are decided: the claim may proceed, and everything else waits for discovery and summary judgment.

  • Decided: one count against two Bloomberg entities (Dkt. 74).
  • Decided: motion to dismiss denied (Dkt. 98).
  • Decided: fair use not decidable on the complaint (Dkt. 98).
  • Decided: schedule extended to 2027 (Dkt. 139).
  • Still open: whether the works were in Books3 and used in training, on evidence.
  • Still open: fair use.
  • Still open: class certification.
  • Still open: damages and any injunction.

Frequently Asked Questions About Huckabee v. Bloomberg

Did Bloomberg lose the case?

No: Bloomberg lost its motion to dismiss on 2025-11-24, which lets the claim proceed. No court has found infringement or rejected fair use as of 2026-11-17.

Has Huckabee v. Bloomberg settled?

No settlement is on the docket entries read through 2026-11-05.

Is BloombergGPT a commercial product?

Bloomberg's answer calls it a research model, and its motion says no public or commercial release was alleged. The court inferred a plausible commercial purpose at the pleading stage only.

What happened to the claims against Meta and Microsoft?

The parties agreed in December 2023 to sever them and transfer them to the Northern District of California (Dkt. 59, 60). The plaintiffs agreed to dismiss EleutherAI without prejudice (Dkt. 59).

When could the court rule on fair use?

Not before summary judgment, and the summary judgment conference is set for 2027-08-24 (Dkt. 139). Fair use was not decided on the complaint.