"Skip the links": Wall Street Journal verklagt KI-Firma Perplexity
Wall Street Journal and New York Post accuse AI company Perplexity AI of large-scale, shameless copyright infringement. Their service harms business.
"Prosperity just around corner"
(Image: Daniel AJ Sokolov)
Perplexity AI, operator of an AI-based "answer machine", is to answer to a US court for infringement of intellectual property rights. The plaintiffs are the parent companies of the Wall Street Journal and its tabloid sister New York Post. They accuse Perplexity of infringing copyright and trademark rights.
Perplexity continuously copies the works of US media and feeds them into a database ("RAG index" for retrieval-augmented generation index, translatable as "rag index" or "cheese sheet index"). A generative language model (LLM) then uses this index to reformulate or summarize content. In some cases, even longer passages are reproduced verbatim.
The plaintiffs see this as a double infringement of copyright: firstly, it is illegal to copy other people's works into the index. Secondly, it is illegal to reproduce excerpts verbatim. According to the statement of claim, Perplexity Pro subscribers even receive the copied excerpts more frequently. The service is not comparable to a traditional search engine, as search engines send users to the websites.
"Skip the links"
Perplexity, on the other hand, advertises its service by saying that users can skip the hyperlinks: "Skip the Links". Instead of being directed to websites by search engines, which enables website operators to earn advertising revenue and offer subscriptions, Perplexity users are supposed to save themselves exactly that. However, this deprives the authors of income, which makes them unhappy. The New York Times, for example, recently sent Perplexity a cease-and-desist letter.
The editions of the language model regularly contain links to the sources; the artificial intelligence therefore at least appears to know where it gets its information from. However, Perplexity also suffers from hallucinations: The language model repeatedly invents content that is not even contained in the cited sources and blames it on them. the plaintiffs are annoyed. They see this as an infringement of their trademark rights.
The lawsuit requests a jury trial and seeks injunctions, the deletion of the database containing the plaintiffs' content, the deletion of all language model editions based on it, liquidated damages of up to USD 150,000 per copyright infringement, as well as disgorgement of profits and treble punitive damages for trademark infringement suffered, plus reimbursement of litigation costs. The case is Dow Jones et NYP Holdings v Perplexity AI and is pending in the US Federal District Court for Southern New York under case number 1:24-cv-07984. heise online has invited Perplexity AI to comment.
Perplexity beckons with revenue share
According to the statement of claim, Perplexity has not yet responded to a request from the plaintiffs in July to enter into negotiations about a license. OpenAI, on the other hand, has reached an agreement with News Corp for license payments of over a quarter of a million US dollars over five years. News Corp is the parent company of Dow Jones and the New York Post.
Perplexity already cooperates with several publishers, including Fortune, Time and Texas Tribune, as well as Spiegel in Germany and Automattic, provider of Wordpress. These are to receive an unspecified share of Perplexity's advertising revenue when their information is used. The Wall Street Journal and the New York Post see this offer as an attempt by Perplexity to unilaterally dictate vague terms after the crime has been committed, which the publishers reject: "When negotiating a valid license for copyrighted material, an infringer does not unilaterally dictate the terms of the license to his victims."
(ds)