Authors Guild v. OpenAI class-action complaint filed (20 September 2023)

On Wednesday 20 September 2023 the Authors Guild and seventeen named author plaintiffs filed a putative class-action complaint in the United States District Court for the Southern District of New York alleging that OpenAI's downloading and ingestion of the plaintiffs' copyrighted novels into the training corpora of its GPT-3.5 and GPT-4 large language models infringed the plaintiffs' copyrights and the copyrights of a proposed class of "fiction writers whose copyrighted works of fiction have been used by OpenAI to train its GPT large language models". The case was styled Authors Guild et al. v. OpenAI Inc. et al. and assigned case number 1:23-cv-08292. It is the opening event of the first major U.S. federal class action by published novelists against an AI developer and, on the public record this corpus tracks, the field-defining U.S. writer-side strategic-litigation moment for the publishing-industry response to generative AI — the publishing-side parallel to the visual-artist-side opening filing eight months earlier in Andersen v. Stability AI.

The complaint

The seventeen named author plaintiffs were David Baldacci, Mary Bly (writing as Eloisa James), Michael Connelly, Sylvia Day, Jonathan Franzen, John Grisham, Elin Hilderbrand, Christina Baker Kline, Maya Shanbhag Lang, Victor LaValle, George R. R. Martin, Jodi Picoult, Douglas Preston, Roxana Robinson, George Saunders, Scott Turow, and Rachel Vail — a cross-genre roster deliberately drawn from the full revenue base of the U.S. trade-fiction market, spanning commercial thriller and mystery (Grisham, Baldacci, Connelly, Preston, Turow), literary fiction (Franzen, Saunders, LaValle, Lang, Robinson), women's fiction and family drama (Picoult, Hilderbrand, Day, Kline, Vail), fantasy and genre (Martin), and the cross-over commercial-literary range (Bly). The original defendants were OpenAI Inc., OpenAI OpCo LLC, OpenAI GP LLC, and OpenAI Holdings LLC; Microsoft Corporation was added as a co-defendant on 4 December 2023 by amended complaint and was not part of the 20 September action.

The pleading advanced three classes of claim — direct, vicarious, and contributory copyright infringement under 17 U.S.C. § 501. The factual core of the complaint was that OpenAI had downloaded the plaintiffs' published novels from pirate ebook shadow-library repositories — Library Genesis, Z-Library, and the Bibliotik tracker — and incorporated those works into the training datasets used to build GPT-3.5 and GPT-4 without permission or compensation; the legal core was that those acts of downloading and ingestion were themselves unauthorised reproductions, and that ChatGPT's output behaviour — its ability to summarise, characterise, and produce continuations of the plaintiffs' fictional worlds in detail — was the operating demonstration that the resulting models retained the plaintiffs' work in a copyright-relevant form. The complaint characterised the conduct as "systematic theft on a mass scale" and asked the court for statutory damages, injunctive relief preventing further training and operation of GPT models on the plaintiffs' works without licences, and the establishment of what the Guild publicly framed as "a fair licensing regime" for AI training on copyrighted books.

Jonathan Franzen and George Saunders were designated as the suit's class representatives. Counsel of record were the class-action firm Lieff Cabraser Heimann & Bernstein — partner Rachel Geman as partner-in-charge — and Cowan, DeBaets, Abrahams & Sheppard LLP, with Scott J. Sholder of CDAS as the publishing-industry-side co-counsel.

Origins: the 18 July 2023 open letter

The 20 September filing was the litigation track of a writer-organising campaign the Authors Guild had been building publicly through the spring and summer of 2023. The Guild had set up its standing AI advocacy programme earlier that year in response to the discovery — through the LLaMA training-corpus disclosure and the subsequent reporting on the Books3 dataset — that books and copyrighted writing had been ingested at scale into the major large-language models. The public-organising arm of the campaign crystallised on 18 July 2023 in the Guild's open letter to AI industry leaders, addressed to the CEOs of OpenAI, Alphabet, Meta, Stability AI, IBM, and Microsoft, signed by 15,000+ writers and bearing the lead-author signatures of Dan Brown, James Patterson, Margaret Atwood, Jodi Picoult, Suzanne Collins, Jonathan Franzen, Roxane Gay, Celeste Ng, Louise Erdrich, Viet Thanh Nguyen, George Saunders, Min Jin Lee, Michael Chabon, Nora Roberts, David Baldacci, and Jesmyn Ward. The letter's three demands — consent before training, compensation for past training uses, and compensation for AI outputs that compete with or substitute for the underlying work — became the substantive frame of the September 2023 complaint, and six of the named litigation plaintiffs (Baldacci, Franzen, Picoult, Saunders, plus the two Guild leaders Maya Shanbhag Lang and Mary Bly) had been signatories to the open letter. The two-month organising-to-litigation pipeline from open letter to filed complaint operationalised what the campaign described from the outset as a paired public-pressure-plus-strategic-litigation track.

Framing on the day of filing

The Authors Guild and the named plaintiffs framed the filing publicly as a campaign-opening action — the litigation vehicle for a longer fight rather than a one-shot legal claim. Authors Guild CEO Mary Rasenberger framed the suit as a stop-the-theft action: "it is imperative that we stop this theft in its tracks or we will destroy our incredible literary culture, which feeds many other creative industries in the U.S. Great books are generally written by those who spend their careers and, indeed, their lives, learning and perfecting their crafts. To preserve our literature, authors must have the ability to control if and how their works are used by generative AI." Rasenberger also publicly characterised the underlying training-data harm as "identity theft on a grand scale" in same-day press, naming the way unconsented ingestion of an author's distinctive prose into a model that can then generate work in the author's voice operates as a category of identity-harm distinct from the conventional copyright claim.

Authors Guild President Maya Shanbhag Lang positioned the lawsuit as the opening of a longer campaign: "the Authors Guild serves to protect the literary landscape and the profession of writing. This case is merely the beginning of our battle to defend authors from theft by OpenAI."

Jonathan Franzen, designated as a class representative, gave the suit its structural framing as a continuation of older industry patterns: "generative AI is a vast new field for Silicon Valley's longstanding exploitation of content providers". George Saunders, the second class representative, anchored the campaign's compensation framing: "writers should be fairly compensated for their work. Fair compensation means that a person's work is valued."

OpenAI's same-day public response reasserted the company's training-data-as-fair-use posture and the company's expressed willingness to engage with creators on rights-management questions — the framing OpenAI would carry through into its motion-to-dismiss pleadings the following spring and which Judge Sidney H. Stein would expressly decline to opine on at the October 2025 motion-to-dismiss stage of the consolidated proceeding.

Significance

The 20 September 2023 filing is the opening event of the Authors Guild v. OpenAI campaign. Three structural features set on the day of filing have continued to shape the case three years on, and have given the wider U.S. publishing sector's AI-policy work its principal organising vehicle.

The first is the seventeen-named-plaintiff frame — a cross-genre roster spanning the U.S. trade-fiction market's full revenue base — which let the suit speak credibly across the disciplinary breadth of the affected writing workforce and gave the press a "Game of Thrones / The Firm" headline framing whose international reach (Variety, Reuters, the BBC, the European publishing press) was structurally larger than any single-discipline filing could have achieved. The 21 November 2023 Alter / Sancton nonfiction-authors filing and the subsequent January–February 2024 consolidation extended that frame across into nonfiction writers, journalists, and historians; the cross-sectoral plaintiff-as-coalition shape of the case was set on 20 September.

The second is the suit's positioning as the writer-side parallel to the visual-artist-side Andersen v. Stability AI filing of 13 January 2023. The two filings — eight months apart, in California and New York respectively, against different AI developers, on different statutory theories, but sharing the underlying "training on copyrighted creative work without consent, credit, or compensation" framing — together constitute the U.S. creative-industry strategic-litigation track on generative-AI training data; the September Authors Guild filing is the moment at which writers caught up to artists on that track and the moment at which the press began to read the two as a single sectoral response. The "consent, credit, and compensation" framing that Matthew Butterick and Joseph Saveri had given the artist-side action carried directly into the writer-side action's "consent / compensation" demand structure, and the Authors Guild's three-pillar open letter the previous July had already operationalised the same framing on the publishing side.

The third is the case's standing as the institutional infrastructure through which the U.S. publishing profession has subsequently engaged federal AI policy. The lawsuit's named-plaintiff roster, public-press footprint, and Authors-Guild-anchored structure gave the wider publishing sector the lobbying vehicle through which it then pursued U.S. Copyright Office ex parte engagement on the 2023–2025 AI study, Capitol Hill engagement on the NO FAKES Act and COPIED Act, and the AI Accountability and Personal Data Protection Act federal-policy push the Guild has carried through the 119th Congress. The substantive content of those tracks descends directly from the consent / compensation / licensing demand structure pleaded into the complaint on the day of filing.