February 2026
Major Development in Generative AI Output-Side Copyright Disputes: New York Federal Court Rules in OpenAI Copyright Litigation That Plaintiffs’ Claims Survive the Pleading Stage
I. Output-Side Infringement Claims Previously Failed to Clear the Procedural Threshold
In copyright disputes involving generative AI, infringement theories are generally divided into “input-side” claims (reproduction of training data) and “output-side” claims (whether the generated results constitute infringing derivative works). In past litigation, defendants (such as OpenAI) frequently argued that plaintiffs failed to specifically identify which AI outputs were “substantially similar” to the original works, or failed to submit concrete examples of infringing outputs. As a result, output-side infringement claims were often dismissed at an early procedural stage.
In April 2025, the U.S. Judicial Panel on Multidistrict Litigation ordered the consolidation of multiple actions [1] , transferring lawsuits originally filed in the Northern District of California—including copyright infringement claims brought by novelists [2] and screenwriters [3] against OpenAI—to the U.S. District Court for the Southern District of New York. These cases were consolidated with actions filed in New York by the Authors Guild [4] and major news media organizations [5] against OpenAI. Given the unprecedented scale of this consolidated proceeding, its subsequent developments were widely expected to become a landmark in generative-AI copyright litigation.
As anticipated, approximately six months after consolidation, the case reached a critical juncture: OpenAI moved to dismiss the plaintiffs’ direct output-side infringement claims, seeking an early procedural dismissal.
II. The Court Found Plaintiffs Had Adequately Pleaded a Prima Facie Case of Infringement Based on Generated Output
In a ruling issued on October 27, 2025, the court denied OpenAI’s motion to dismiss, holding that the plaintiffs had sufficiently established a prima facie case of copyright infringement:
1. Preliminary showing of actual copying:
The court found that the plaintiffs had plausibly alleged that OpenAI had access to their works (used in training the foundation models) and that ChatGPT’s generated outputs were indeed derived from those works.
2. Substantial similarity:
Applying the “more discerning observer test,” the court analyzed the ChatGPT output examples submitted by the plaintiffs, including detailed plot summaries of A Song of Ice and Fire and unauthorized sequel drafts. The court noted that the generated content went beyond mere factual summaries and instead engaged in extensive “parroting” of the original works’ core copyright-protected elements—such as distinctive character traits, plot development, setting, and the overall look and feel of the works—such that a reasonable jury could find substantial similarity.
III. Whether the Output Actually Constitutes Copyright Infringement Hinges on the Fair Use Analysis, Pending Further Substantive Trial
Although the ruling represents a significant victory for the plaintiffs, the court emphasized that clearing the pleading threshold does not equate to a finding of infringement. The judge expressly stated that the court took no position on the fair use defense at this stage. Fair use is a highly fact-intensive inquiry and is typically addressed at the summary judgment stage, after discovery has produced further evidence regarding the purpose and character of the use, its transformative nature, and its impact on the market for the original works.
Accordingly, the ultimate determination of whether the AI-generated outputs infringe copyright remains to be resolved in subsequent substantive proceedings.
[1] In re OpenAI, Inc., Copyright Infringement Litigation, No. 1:25-md-03143 (S.D.N.Y. centralized Apr. 3, 2025).
[2] Tremblay et al. v. OpenAI, Inc. et al., No. 3:23-cv-03223 (N.D. Cal. filed June 28, 2023).
[3] Silverman v. OpenAI, Inc., No. 3:23-cv-03416 (N.D. Cal. filed July 7, 2023).
[4] Authors Guild v. OpenAI, Inc., No. 1:23-cv-08292 (S.D.N.Y. filed Sept. 19, 2023).
[5] The New York Times Co. v. Microsoft Corp., et al., No. 1:23-cv-11195 (S.D.N.Y. filed Dec. 27, 2023).
In copyright disputes involving generative AI, infringement theories are generally divided into “input-side” claims (reproduction of training data) and “output-side” claims (whether the generated results constitute infringing derivative works). In past litigation, defendants (such as OpenAI) frequently argued that plaintiffs failed to specifically identify which AI outputs were “substantially similar” to the original works, or failed to submit concrete examples of infringing outputs. As a result, output-side infringement claims were often dismissed at an early procedural stage.
In April 2025, the U.S. Judicial Panel on Multidistrict Litigation ordered the consolidation of multiple actions [1] , transferring lawsuits originally filed in the Northern District of California—including copyright infringement claims brought by novelists [2] and screenwriters [3] against OpenAI—to the U.S. District Court for the Southern District of New York. These cases were consolidated with actions filed in New York by the Authors Guild [4] and major news media organizations [5] against OpenAI. Given the unprecedented scale of this consolidated proceeding, its subsequent developments were widely expected to become a landmark in generative-AI copyright litigation.
As anticipated, approximately six months after consolidation, the case reached a critical juncture: OpenAI moved to dismiss the plaintiffs’ direct output-side infringement claims, seeking an early procedural dismissal.
II. The Court Found Plaintiffs Had Adequately Pleaded a Prima Facie Case of Infringement Based on Generated Output
In a ruling issued on October 27, 2025, the court denied OpenAI’s motion to dismiss, holding that the plaintiffs had sufficiently established a prima facie case of copyright infringement:
1. Preliminary showing of actual copying:
The court found that the plaintiffs had plausibly alleged that OpenAI had access to their works (used in training the foundation models) and that ChatGPT’s generated outputs were indeed derived from those works.
2. Substantial similarity:
Applying the “more discerning observer test,” the court analyzed the ChatGPT output examples submitted by the plaintiffs, including detailed plot summaries of A Song of Ice and Fire and unauthorized sequel drafts. The court noted that the generated content went beyond mere factual summaries and instead engaged in extensive “parroting” of the original works’ core copyright-protected elements—such as distinctive character traits, plot development, setting, and the overall look and feel of the works—such that a reasonable jury could find substantial similarity.
III. Whether the Output Actually Constitutes Copyright Infringement Hinges on the Fair Use Analysis, Pending Further Substantive Trial
Although the ruling represents a significant victory for the plaintiffs, the court emphasized that clearing the pleading threshold does not equate to a finding of infringement. The judge expressly stated that the court took no position on the fair use defense at this stage. Fair use is a highly fact-intensive inquiry and is typically addressed at the summary judgment stage, after discovery has produced further evidence regarding the purpose and character of the use, its transformative nature, and its impact on the market for the original works.
Accordingly, the ultimate determination of whether the AI-generated outputs infringe copyright remains to be resolved in subsequent substantive proceedings.
[1] In re OpenAI, Inc., Copyright Infringement Litigation, No. 1:25-md-03143 (S.D.N.Y. centralized Apr. 3, 2025).
[2] Tremblay et al. v. OpenAI, Inc. et al., No. 3:23-cv-03223 (N.D. Cal. filed June 28, 2023).
[3] Silverman v. OpenAI, Inc., No. 3:23-cv-03416 (N.D. Cal. filed July 7, 2023).
[4] Authors Guild v. OpenAI, Inc., No. 1:23-cv-08292 (S.D.N.Y. filed Sept. 19, 2023).
[5] The New York Times Co. v. Microsoft Corp., et al., No. 1:23-cv-11195 (S.D.N.Y. filed Dec. 27, 2023).
