IP Law Daily, COPYRIGHT NEWS: Entertainment metadata creator sues OpenAI, alleges unauthorized use of copyrighted works for AI training, (Mar 12, 2026)

By Robert Margolis, J.D.

Gracenote alleges OpenAI’s unauthorized uses harm Gracenote both in the market for media metadata content and delivery, as well as in the market for licensing media metadata content for AI training.

In the latest salvo fired by owners of copyrighted content against allegedly unauthorized use of copyrighted works in the training of AI models, media metadata creator Gracenote Media Services has filed a four-count complaint against OpenAI Foundation (formerly known as OpenAI, Inc.) and several affiliates in the United States District Court for the Southern District of New York. Gracenote alleges OpenAI has used Gracenote’s metadata without authorization to train its ChatGPT AI products. Gracenote has brought claims for copyright infringement, vicarious copyright infringement, contributory copyright infringement, and unjust enrichment. The crux of the complaint is that to build its ChatGPT AI products, OpenAI has (1) copied Gracenote’s copyrighted metadata without authorization, (2) used it to train, develop, operate, and improve its ChatGPT products, (3) used it to deliver infringing content to users of OpenAI products and services, (4) undermined Gracenote’s role in the markets for metadata itself and for the use of metadata to train AI products, and (5) unjustly enriched itself by doing so (Gracenote Media Services, LLC v. OpenAI Foundation, No. 1:26-cv-01947 (S.D.N.Y. Mar. 10, 2026)).

Copyrighted materials. Gracenote alleges it has developed for over half a century a series of metadata products, which it calls the “Gracenote Programs Database and its data (collectively ‘Gracenote Data’),” for helping media consumers and media content distributors “navigate and discover media content.” The Gracenote Data is metadata used for identifying TV shows, movies, and other media content. The Gracenote Data “includes millions upon millions of narrative descriptions, original video descriptors, unique identifiers, and other program elements” that Gracenote editors have written, edited, selected, and arranged, and which Gracenote alleges is “tremendously valuable.” The complaint details the “painstaking” work by its editors to “source, ingest, aggregate, research, edit, write, curate, and link content” for the Gracenote Programs Database. This includes the creation of “program descriptions, genre selections, and proprietary video descriptors” (“Descriptive Records”) that capture “detailed program attributes such as mood, theme, scenario, setting, and subject.” These Descriptive Records are proprietary as are their editors’ determinations of how to link and arrange categories of content in ways helpful to its customers. The complaint also describes how Gracenote has created a “proprietary identification system,” through the assignment of a unique 14-character alphanumeric identifier (“TMSID”) to each program (such as TV shows and movies).

Customers of this data include large providers of media content distributors (such as cable, satellite, and streaming services), who license the Gracenote Data for use by their own customers. In addition, Gracenote alleges that the Gracenote Data is “highly valuable” to AI and machine learning (ML) providers, who license the data to train and improve their models.

Gracenote alleges it has registered the Gracenote Programs Database with the United States Copyright Office to protect its copyrights in the Gracenote Data, and that it owns all right, title, and interest in and to these copyrights.

Allegedly infringing conduct. Gracenote alleges that to train the large-language models (LLMs) incorporated in several ChatGPT products, OpenAI has without authority “improperly obtained, copied, and used” the Gracenote Data. It alleges “on information and belief” that in training certain of its models, OpenAI used the “Common Crawl” dataset, essentially a “copy of the internet,” which includes numerous web domains containing Gracenote Data, including tvlistings.gracenote.com. This use violates Gracenote’s Terms of Use, applicable to its website and all related web pages, and which identifies and asserts copyright ownership over all content.

The complaint includes examples of how OpenAI’s ChatGPT-4 models can recall TMSIDs associated with specific programs, which could only be done by copying the Gracenote Data in which those TMSIDs are found. It also provides examples of how several Chat-GPT models “memorize” Descriptive Records, and then “regurgitate” them verbatim when prompted to do so.

Use of the Gracenote Data without a license is unauthorized. Gracenote forbids its clients from providing the Gracenote Programs Database or its components to third parties, except in limited circumstances not applicable to OpenAI’s uses, according to Gracenote. It alleges that OpenAI has never sought to obtain a license to use Gracenote Data, and that the resulting unauthorized use by OpenAI is willful.

Gracenote alleges that OpenAI’s unauthorized uses of the Gracenote Data harms it in the market for media metadata content and discovery, as well as in the market for media metadata content licensing to train AI models.

Infringement claims. Gracenote alleges that by building training datasets that include unauthorized copies of Gracenote Data, and by storing and processing the training datasets containing such unauthorized copies, OpenAI has directly infringed Gracenote’s copyrights. It alleges “on information and belief” that this conduct is “willful.”

Gracenote also alleges that several OpenAI entities are liable for “vicarious copyright infringement.” Further, to the extent end users may be liable as direct infringers based on the outputs generated by OpenAI’s ChatGPT products, Gracenote alleges OpenAI is liable for contributory infringement. Finally, Gracenote asserts a claim for unjust enrichment, based on OpenAI’s use and reliance on the Gracenote Data to train its ChatGPT models, from which it has benefitted without compensating Gracenote for the value of this use.

Relief requested. Gracenote seeks statutory damages, compensatory damages, restitution, disgorgement, and any other relief available in law or equity. It also seeks a permanent injunction against further infringing conduct as alleged in the complaint, an order under 17 U.S.C. § 503(b) of all OpenAI’s models and training sets that incorporate Gracenote Data, as well as attorney fees.

The Case is No. 1:26-cv-01947.

Attorneys: Elisha Brandis Barron (Susman Godfrey LLP) for Gracenote Media Services, LLC.

Companies: Gracenote Media Services, LLC; OpenAI Foundation f/k/a OpenAI, Inc.

News: Copyright NewYorkNews AINews TechnologyInternet