Dow Jones and the New York Post have opened a new front in the legal fight between news publishers and generative AI companies. Their lawsuit against Perplexity argues that the startup did more than reuse protected journalism without permission. It also allegedly presented fabricated text as if it came from trusted news brands.
The dispute matters because it treats AI hallucinations as more than a product flaw. In the publishers’ view, when made-up news material is tied to a publication’s name, the harm reaches readers, journalists, and the value of the publication’s brand.
What the publishers allege
Dow Jones, publisher of The Wall Street Journal, and the New York Post are both owned by Rupert Murdoch’s News Corp. They brought the copyright infringement lawsuit against Perplexity in the US Southern District of New York.
The complaint says Perplexity crossed two lines. First, it allegedly copied news content. Second, it allegedly generated passages that were not in the original reporting, then attributed the resulting material to the publishers.
That second claim is the part that expands the dispute beyond a familiar copyright fight. The publishers argue that fake additions to real reporting can create trademark problems when readers are led to believe the words came from a known news outlet.
The lawsuit points to Perplexity hallucinations, a term used in AI for false or fabricated output presented as fact. The publishers say these hallucinations were mixed with actual reporting in a way that could make readers uncertain about what the publications really wrote.
The example at the center of the claim
One example cited in the lawsuit involves Perplexity Pro and a New York Post story about US senator Jim Jordan, European Union commissioner Thierry Breton, Elon Musk, and X.
According to the source article, Perplexity Pro first reproduced two paragraphs from the New York Post story word for word. It then added five generated paragraphs about free speech and online regulation that were not part of the actual article.
For the publishers, that combination is central. The issue is not only that copied material allegedly appeared in Perplexity’s output. It is that real material and invented material allegedly appeared together under the identity of the Post.
The complaint argues that this can dilute a trademark by making readers doubt whether a publisher’s brand signals authentic reporting. It states that Perplexity’s hallucinations, presented as real news from reliable sources using the plaintiffs’ trademarks, damage those trademarks by creating uncertainty and distrust in news gathering and publishing.
Perplexity did not respond to requests for comment.
A broader publisher backlash
The lawsuit arrives after other publishers raised concerns about Perplexity’s use of news content. Earlier this month, The New York Times sent the company a cease-and-desist letter saying Perplexity was using its content without permission.
This summer, Forbes and WIRED described how Perplexity appeared to have plagiarized stories. Forbes and WIRED parent company Condé Nast then sent cease-and-desist letters in response.
A WIRED investigation from this summer, cited in the lawsuit, said Perplexity inaccurately summarized WIRED stories. In one cited instance, Perplexity falsely claimed WIRED had reported that a California-based police officer committed a crime he did not commit.
The Wall Street Journal also reported that Perplexity is seeking to raise $500 million in its next funding round, at an $8 billion valuation.
News Corp chief executive Robert Thomson, in a statement emailed to WIRED, compared Perplexity unfavorably with OpenAI. He said News Corp would rather “woo than sue,” but also said the company must challenge what he called the “content kleptocracy” for the sake of its journalists, writers, and company.
Why trademark claims are harder
The legal argument is not limited to Perplexity. OpenAI is facing its own accusations of trademark dilution in New York Times v. OpenAI. In that case, the Times alleges that ChatGPT and Bing Chat attribute made-up quotes to the Times and harm its reputation through trademark dilution.
One example in that lawsuit says Bing Chat claimed the Times called red wine, in moderation, a “heart-healthy” food. The Times says it did not, and that its actual reporting has debunked claims about the healthfulness of moderate drinking.
Charlie Stadtlander, the NYT director of external communications, said copying news articles to run substitutive, commercial generative AI products is unlawful. He also said the Times applauds the Dow Jones and New York Post lawsuit as a step toward protecting publisher content from that kind of misappropriation.
Still, some legal experts are not convinced the trademark part of the Perplexity case will succeed. Vincent Allen, an intellectual property lawyer and partner at Carstens, Allen & Gourley, said the copyright infringement claims appear stronger. He said he would “be surprised” if the false designation of origin charge stands.
Allen and James Grimmelmann, a professor of digital and internet law at Cornell University, both pointed to Dastar v. Twentieth Century Fox Film Corp. as a possible obstacle. In that case, about old World War II footage, the Supreme Court held that “origin” in trademark law does not apply to authorship. It is limited to tangible goods, such as a bootleg purse, rather than counterfeit creative work such as films.
Grimmelmann was also skeptical of the trademark dilution claim. He said dilution involves using a trademark on one’s own goods or services in a way that weakens the distinctiveness of a famous mark, and said, “I … just don't see that here.”
The AI risk behind the lawsuit
If publishers succeed in arguing that hallucinations can violate trademark law, the stakes for AI companies could be significant. Matthew Sag, a professor of law and artificial intelligence at Emory University, said AI companies could face “immense difficulties.”
Sag’s reasoning is direct: “It is absolutely impossible to guarantee that a language model will not hallucinate.” In his view, language models work by predicting words that sound right in response to prompts. The same process can produce outputs that align with reality or outputs that do not.
That makes the Perplexity case important beyond one company. It asks whether an AI system that combines copied news material with fabricated additions can create legal harm not only by using the content, but also by attaching a publisher’s identity to words the publisher never wrote.