Authors Guild et al. v. OpenAI Inc. et al., No. 1:23-cv-08292 (S.D.N.Y.), is a putative class action brought by the Authors Guild and a group of named novelists alleging that OpenAI reproduced their copyrighted books without licence, used those reproductions to train its large language models, and produced infringing outputs through ChatGPT. It was filed on September 19, 2023 with John Grisham, Jodi Picoult, David Baldacci, George R.R. Martin, and thirteen other authors as named plaintiffs (Source: authorsguild.org).
The case is now one component of a consolidated multidistrict proceeding rather than a standalone action, and its October 2025 ruling on output-based infringement is the point at which it diverges from the training-side reasoning that governs Bartz v. Anthropic. The wider set of suits is tracked at AI Copyright Litigation — Analysis.
Case information
| Field | Detail |
|---|---|
| Lead docket | 1:23-cv-08292 (S.D.N.Y.) |
| MDL | In re: OpenAI, Inc. Copyright Infringement Litigation, MDL No. 3143 (1:25-md-03143) |
| Judge | Sidney H. Stein |
| Defendants | OpenAI entities; Microsoft Corporation |
| Filed | September 19, 2023 |
| Status | Active |
Claims
The Consolidated Class Action Complaint alleges that OpenAI and Microsoft infringed the plaintiffs' copyrights in their books in three ways: by downloading and reproducing the works; by using those reproduced works to train OpenAI's large language models; and by creating infringing works in the outputs of OpenAI's products, including ChatGPT (Source: law.justia.com).
The third theory — output infringement — distinguishes the case from the acquisition-and-training claims that dominate the rest of the copyright docket. The plaintiffs allege that ChatGPT generates derivative works including detailed plot summaries that reproduce protectable elements of their novels.
Procedural history
OpenAI and Microsoft each answered on February 16, 2024. A consolidated complaint was filed on February 2, 2024, combining the Authors Guild action with Alter v. OpenAI, Microsoft (formerly Sancton v. OpenAI, Microsoft). Basbanes v. Microsoft, filed January 5, 2024, was consolidated for pretrial purposes and subsequently stayed pending a joint submission within two weeks of a ruling on class certification.
On April 3, 2025 these class actions and other related actions against OpenAI and Microsoft were centralized in a multidistrict litigation before Judge Stein in the Southern District of New York. The Northern District of California cases consolidated as In re OpenAI ChatGPT Litigation — comprising Tremblay v. OpenAI, Silverman v. OpenAI, and Chabon v. OpenAI — were transferred into the same MDL (Source: chatgptiseatingtheworld.com). The transferred California actions are separately covered at Tremblay v. OpenAI.
Oral argument on OpenAI's motion to dismiss the Consolidated Class Action Complaint was held October 8, 2025.
The October 2025 output ruling
On October 27, 2025 the court denied OpenAI's motion to dismiss the class plaintiffs' claim of direct copyright infringement based on ChatGPT's outputs, holding that the allegations satisfy the elements of a prima facie infringement claim "as to at least some outputs of ChatGPT."
The court addressed two threshold questions before reaching substantial similarity. It declined to hold OpenAI to a January 19, 2024 stipulation not to seek dismissal of claims then pending in the Authors Guild action, reasoning that the circumstances, case schedule, and identity of the plaintiffs had changed significantly after centralization in the MDL and consolidation of the class actions. It also held that the outputs the plaintiffs submitted with their opposition brief were incorporated into the Consolidated Class Action Complaint by reference and were integral to the output-based allegations, because the complaint makes "clear, definite and substantial reference[s]" to them.
On the merits of the motion, the court applied the "more discerning observer" test, asking whether substantial similarity exists "between those elements, and only those elements, that provide copyrightability to the allegedly infringed [work]." It distinguished its own earlier determination that summaries of news articles were not substantially similar to the underlying copyrighted works, finding that "the outputs submitted by plaintiffs in this action incorporate such copyrightable elements of plaintiffs' original works as plot, setting, and characters." Because it is sufficient at the pleading stage that plaintiffs have alleged some outputs a reasonable jury could find substantially similar, the motion was denied (Source: law.justia.com).
Current status
The case is active in the MDL, with summary-judgment replies scheduled for October 16, 2026 and extensive discovery disputes on the docket (Source: chatgptiseatingtheworld.com). Class certification had not been resolved as of the last status check; the Basbanes stay is keyed to that ruling.
The Authors Guild's role in the wider dispute extends beyond this docket. It has advanced collective-licensing proposals discussed at AI Content Licensing, and it characterizes the $1.5 billion settlement in Bartz v. Anthropic as the largest copyright settlement in United States history. The suit is also the predicate for the shareholder-disclosure theory in the Microsoft derivative complaint summarized at Anderson v. Microsoft — Shareholder Derivative Complaint (June 2026), which describes it as "the first major class action against Microsoft" of its kind.
The doctrinal framework the case operates within — the division between ingestion, output, authorship, and licensing questions — is set out at AI Copyright.
Relationships
- instance-of: AI Copyright
- related: AI Copyright Litigation — Analysis
- related: Bartz v. Anthropic
- related: NYT v. Microsoft, OpenAI et al.
- related: Tremblay v. OpenAI
- related: AI Content Licensing
- supports: Anderson v. Microsoft — Shareholder Derivative Complaint (June 2026)
- litigates: OpenAI
Open questions
- Whether the October 2025 pleading-stage holding on output similarity survives summary judgment, where the works themselves rather than the allegations control, is unresolved.
- Whether a class can be certified across authors whose works differ in the degree to which model outputs reproduce protectable elements has not been decided.