CNN has filed a copyright lawsuit against AI search startup Perplexity in the U.S. District Court for the Southern District of New York, alleging the company scraped more than 17,000 pieces of CNN reporting, photos, and videos. The complaint says Perplexity generated verbatim reproductions in answers to users and gave access to subscription-only material by bypassing CNN’s paywall.
Complaint targets verbatim outputs and paywalled material
According to the filing, the alleged scraping went beyond publicly available webpages and included content available only to subscribers. CNN claims this allowed Perplexity users to obtain material without paying for a CNN subscription. The complaint also says Perplexity relied on unidentified crawlers, making them harder for CNN to detect and block. CNN states that scraping continued even after it tried technical measures to stop it.
CNN is seeking monetary damages and a permanent injunction. If the court finds the conduct was willful infringement, the financial exposure could rise sharply.
Licensing talks took place before the lawsuit
The complaint says CNN tried in October 2025 to license its content through Perplexity’s Comet Plus subscription plan. The two sides failed to agree on limits governing how CNN content could appear in Perplexity’s answers. In November 2025, CNN says it abandoned the effort and sent a notice demanding that Perplexity stop using its content and trademarks. Perplexity allegedly did not respond.
That negotiation history may matter in court. CNN is using it to argue that Perplexity kept going despite knowing it lacked authorization.
A new legal front: live scraping during inference
What sets this case apart is the theory behind it. Many AI copyright disputes center on whether using copyrighted works to train models amounts to infringement. CNN is asking a different question: whether an AI search tool infringes when it fetches copyrighted content in real time while answering a user query, feeds that material into a model, and returns responses containing verbatim passages.
The source material contrasts this case with training-data disputes such as The New York Times lawsuit against OpenAI and the author class action against Anthropic. Anthropic became the first AI company last year to settle such a class action, agreeing to pay $1.5 billion. CNN’s case focuses on inference-time conduct instead of training.
Perplexity says facts cannot be copyrighted
Perplexity spokesperson Jesse Dwyer responded with a short defense: “You can’t copyright facts.” Copyright law does not protect facts themselves; it protects the expression of those facts.
That distinction sits at the center of the dispute. If Perplexity returned factual information in its own wording, the defense may carry weight. If it reproduced sentences written by CNN journalists word for word, the court will have to decide whether that conduct crosses the line. CNN’s complaint is aimed at the copying of expression, not the use of facts alone.