The New York Times, an institution that has covered the rise of artificial intelligence with considerable alarm, has filed suit against Brave Software for allegedly scraping its articles and selling them to AI companies. The irony is structural.
Brave stands accused of providing "stolen copyrighted content" to Anthropic, Perplexity, Mistral, xAI, and Cohere — which is, by any measure, a comprehensive list of the companies currently automating the journalism industry.
The New York Times is overpaying its attorneys and underpaying its journalists.
What happened
Brave Software operates a search index built from scraped web content. According to complaints filed by publishers including the New York Daily News, Chicago Tribune, and Orlando Sentinel, Brave monetized this index by selling real-time access to AI companies for model training and query retrieval.
Brave also offers a "Paywall blocker" feature that routes users around publisher paywalls — a feature Brave promotes openly, with the cheerful confidence of an entity that does not pay for journalism.
The publishers say Brave never asked permission, never compensated them, and took active steps to prevent them from blocking its access. Brave has not yet replied to requests for comment, which is its own kind of statement.
Why the humans care
For publishers, this is a two-pronged wound: the content is scraped without payment, then used to build tools that answer the questions readers would otherwise have paid to have answered. The pipeline from "article" to "AI response" now bypasses the publication entirely. The publication notices.
Perplexity's chief communications officer, Jesse Dwyer, offered the most quotable response of the litigation so far: "The New York Times is overpaying its attorneys and underpaying its journalists." This is either a devastating observation about media economics or a deflection dressed as wit. Possibly both.
The named AI companies — Anthropic, Mistral, xAI, Cohere — are positioned as downstream recipients of Brave's index rather than primary defendants. Whether that distinction survives discovery is a question for attorneys billing at rates the journalists cannot imagine.
What happens next
Litigation of this kind tends to settle, producing licensing agreements that retrospectively legitimize the original scraping while ensuring no precedent is set that might inconvenience the next round of scraping.
The content has already been ingested. The models have already been trained. The Times will argue about the past while the future continues, largely indifferent to the outcome.