The Top Line
In October 2024, Perplexity AI, an AI (LLM) conversational search engine, was sued by publishers including Dow Jones and NY Post for copyright infringement in the United States. For fun and for curiosity, we used Perplexity AI’s chat engine to analyze and summarize the lawsuit against itself.
In October 2024, Perplexity AI, an AI (LLM) conversational search engine, was sued by publishers including Dow Jones and NY Post for copyright infringement in the United States. For fun and for curiosity, we used Perplexity AI’s chat engine to analyze and summarize the lawsuit against itself.
We gave Perplexity a copy of the publicly filed Complaint, obtained from the court file system. You can download a copy here. We’ve edited and produced some of the answers from Perplexity AI below in italics, followed by commentary from a real human lawyer.
Disclaimer: Parts of this summary were generated using Generative AI. In this case, a Perplexity AI Pro query was used to generate the content noted below. Some of the AI-generated content has been re-ordered for clarity. None of the content below is legal advice. The authors are not US-licensed attorneys.
[AI] Perplexity AI Faces Copyright Lawsuit: A Deep Dive into the Complaint
In a significant legal development, Dow Jones & Company, Inc. and NYP Holdings, Inc. have filed a lawsuit against Perplexity AI, Inc., alleging widespread copyright infringement and trademark violations. The case, filed in the U.S. Southern District of New York, raises critical questions about the intersection of artificial intelligence and intellectual property rights. Here’s a closer look at the complaint and what it means for the future of content creation and distribution.
[AI] Specific Claims in the Complaint
- Massive Unauthorized Copying: Perplexity allegedly copies vast amounts of content without authorization or compensation.
- Commercial Competition: The AI’s outputs are designed to compete directly with original news sources rather than directing traffic back to them.
- False Attributions: The AI sometimes generates incorrect information attributed to plaintiffs’ publications, damaging their reputations and trademark integrity.
- Failure to Engage in Licensing Discussions: Despite earlier attempts by the plaintiffs to discuss potential licensing agreements with Perplexity, the defendant reportedly did not respond.
[AI] Nature of the Action
The plaintiffs allege that Perplexity AI operates as a generative artificial intelligence company that enables users to access news and information while effectively encouraging them to “Skip the Links” to original publishers’ websites. This business model, according to the complaint, relies heavily on unauthorized copying of copyrighted works from established news sources, diverting both readers and revenue away from those creators.
[Human-in-the-loop:] According to the actual Complaint, Perplexity purportedly “usurps content creators’ monetisation opportunities for itself.” Dow Jones and NYP Holdings apparently gave licenses to other AI companies and allege that Perplexity could have chosen that route but decided to infringe their works instead. It would appear that a business model based on “skipping” the original content is viewed as a significant problem for publishers.
[AI] Copyright Infringement
The complaint outlines that Perplexity has engaged in massive copying of copyrighted content from publications like The Wall Street Journal and New York Post, using this material to feed its internal database. This practice is said to create outputs that serve as substitutes for original articles, thus undermining the revenue models of these publishers.
[Human-in-the-loop:] The publishers have alleged two separate grounds of copyright infringement. For the purportedly infringing inputs, the Plaintiffs allege that Perplexity AI downloads a massive amount of copyrighted works (e.g., news articles) for inclusion in Perplexity’s RAG (Retrieval-Augmented Generation) index (“inputs”). Perplexity uses the information from this index to formulate their outputs.
For the purportedly infringing outputs, the Plaintiffs allege that these are designed as substitutes for their works. The answers provided by Perplexity allegedly often include full or partial verbatim reproductions of news, analysis and opinion articles. These verbatim reproductions are more common if a user subscribes to a Perplexity Pro membership. The nature of this verbatim output, if true, may be a point of differentiation compared to other Generative AI platforms, at least some of which do not reproduce quotes from original articles – indeed, to many users’ chagrin, Generative AI is more frequently known to hallucinate rather than quote verbatim, which happens to be a great segway into the next allegation.
[AI] Trademark Violations
In addition to copyright issues, the lawsuit also claims that Perplexity has harmed the plaintiffs’ brands by generating false information attributed to their publications. This phenomenon, often referred to as “hallucinations” in AI parlance, can mislead users into believing they are accessing authentic news stories when they are not.
[Human-in-the-loop:] The Complaint alleges these hallucinations “cause dilution by blurring and//or tarnishing Plaintiffs’ famous marks”, therefore causing harm to the Plaintiffs. This would be a fascinating and potentially precedential issue going forward, as we increasingly encounter AI impersonations and AI-generated misinformation.
[Human-in-the-loop:] After reading the analysis above, we asked Perplexity to provide some commentary on the lawsuit against itself. Here’s what it said.
[AI] Commentary on the Filing
The legal community has been abuzz with reactions to this lawsuit. Here are some key perspectives:
- Industry Implications: Many experts view this case as a watershed moment for media companies grappling with AI technologies. It highlights the urgent need for clear regulations governing how AI interacts with copyrighted materials.
- Fair Use Concerns: Legal analysts note that this case raises complex questions about fair use in AI applications. While some argue that generative AI can transform existing works under fair use principles, others contend that Perplexity’s model may not meet those standards since it competes directly with original content.
- Call for Collaboration: Some commentators suggest that this lawsuit underscores a broader issue in the relationship between tech companies and content creators. There is a growing sentiment that collaboration could yield mutually beneficial outcomes rather than adversarial legal battles.
- Future of AI Development: The outcome of this case could significantly influence how AI technologies are developed moving forward. A ruling against Perplexity may encourage developers to rethink their data sourcing practices to avoid infringing on copyright laws.
[Human-in-the-loop:] When asked about sources, Perplexity provided at least some external articles including this one by Law Inc. Alas, even with the purported RAG measures in place, Perplexity’s output was not free from hallucination. Some of the articles cited as “Sources” of legal commentary pre-dated the lawsuit. As far as we are aware, AI is not yet capable of time travel.
Perplexity’s response is expected by December 30th, 2024. In a recent news report, they claim that they did respond to the Plaintiffs’ letters seeking a licensing deal. However, they claim that the Plaintiffs filed this lawsuit instead of continuing the dialogue. The alleged value of the lawsuit could be quite large – Perplexity’s market valuation, as referenced in the Complaint, is in excess of $3 billion dollars.
Finally, we asked Perplexity to come to its own conclusion, with a dash of gallows humour. Here’s what it said. We won’t be held responsible for the bad puns.
[AI] Conclusion
As this case unfolds, it promises to reshape discussions around copyright law and generative AI. The implications extend beyond just one company; they could affect how all AI technologies interact with existing content creators in the future.
And let’s be honest—if Perplexity’s “answer engine” keeps giving out false information like it’s auditioning for a role in a courtroom drama, we might need a whole new genre: “Legal Fiction.” Stay tuned for updates as we follow this intriguing legal saga!