Saturday, Oct 31 | --:--
Times of AI

Regions

China Europe India Japan Taiwan United Kingdom United States

Platforms

Alibaba Qwen Amazon Anthropic DeepMind Google Groq Meta Microsoft MiniMax Mistral Moonshot AI NVIDIA OpenAI PrismML Replit Thinking Machines xAI

Sections

Business Engineering Enterprise Frontier Labs Hardware Industry Infrastructure Models Policy Products Research Safety Security Work All articles →
Back to home
Policy Legal United States OpenAI Microsoft

Wikimedia Asks the Court in the OpenAI Copyright Case to Judge Training and AI Grounding Separately, as Publishers and OpenAI's Allies File Briefs

Friend-of-the-court briefs landed on October 30 in In re: OpenAI, Inc. Copyright Infringement Litigation, the combined New York Times and publisher suits now at summary judgment. The Wikimedia Foundation, backing neither side, argued that training on copyrighted work can be fair use while retrieval-augmented generation and similar grounding should honor licenses. The News/Media Alliance attacked both uses. OpenAI published excerpts from briefs supporting fair use for training.

Times of AI Desk 4 min read New York, New York View as Markdown
Tall stacks of string-tied papers rest on the bench of an empty wood-panelled courtroom, an empty high-backed chair behind and rows of empty pews in front.

One of Friday's briefs in the New York copyright fight over ChatGPT came from a party that wants neither side to win outright, and it reframes the question. The Wikimedia Foundation asked the court on October 30 to treat training a model and using live web content to ground a chatbot's answers as two different questions. Its brief argues the first can be fair use and the second should respect the source's license. The publishers' brief treats both uses as substitutes for journalism, and the excerpts OpenAI published from its supporters' briefs focus on training.

What is being decided#

The briefs were filed in In re: OpenAI, Inc. Copyright Infringement Litigation, the multidistrict case in the Southern District of New York that combines suits against OpenAI and Microsoft, including The New York Times' 2023 case. Wikimedia describes it as 12 combined lawsuits. The News/Media Alliance said the publisher suits are now at summary judgment before Judge Stein, and called it the first time a US court will decide summary judgment on OpenAI's conduct or on real-time use of news content.

Wikimedia: split the question#

Wikimedia argues that using copyrighted content to train a model should be fair if the model is something new and different from the original sources, with no licensing requirement. For retrieval-augmented generation and newer techniques that serve a similar grounding purpose, it argues the use should not be considered fair, and providers should follow each publisher's licensing terms. For Wikipedia that means CC BY-SA 4.0, which requires credit and that adaptations stay under the same license. It also says scraping of its projects should stay within its public API rate limits or go through its Enterprise API.

The Verge reported that Wikimedia sides with neither party and quoted the brief's concern that "the parties' positions in this case push for overbroad and simplified applications of the fair use doctrine." The filing follows Wikimedia's allegation earlier this month that agents it believes were OpenAI's edited its wikis and made millions of API requests.

Publishers: both uses substitute for news#

The News/Media Alliance, joined by the National Newspaper Association, the National Press Photographers Association and ProPublica, urged the court to side with the publishers. Its brief argues that copying news to train models and then to answer users through RAG is not transformative because the products substitute for the original journalism. The Alliance also cited research it says shows ChatGPT sends 96% less referral traffic to publishers than traditional Google search. "This case is existential for the news media industry," said Danielle Coffey, its chief executive. Referral traffic is already a live business problem elsewhere: Reddit said this week that its Google search referrals had fallen.

OpenAI's side: training is the case#

OpenAI on the same day published excerpts from briefs supporting fair use for AI training, from groups including Authors Alliance, Public Knowledge, Engine and Developers Alliance, and a coalition of the American Council of the Blind, library associations, the Consumer Technology Association and the Business Software Alliance. Its post also quotes the US government, whose statement says the Times "seeks to narrow fair-use doctrine to exclude the training of OpenAI's large language models." OpenAI wrote that the Times is pursuing "an undeserved payday." None of the excerpts addresses retrieval-time use of news, the use Wikimedia singles out.

Limits#

  • We read the organizations' announcements and OpenAI's excerpts, not the full briefs on the court docket.
  • The 96% referral-traffic figure is the News/Media Alliance's characterization of outside research; we have not reviewed that research.
  • OpenAI's post does not date each brief. It quotes the US government's filing without a date; that filing may not be new.
  • Amicus briefs do not bind the court, and no summary-judgment ruling has been issued.

Sources#

Prior Coverage

Earlier Times of AI reporting on this thread.

Related Coverage

Scroll to continue reading