The New York Times and three Tribune-group papers sue Brave, targeting the search API that feeds AI companies

On November 2, The New York Times and a second group of papers (the New York Daily News, the Chicago Tribune and the Orlando Sentinel) filed separate copyright suits against Brave Software in the Southern District of New York: No. 26-cv-09549 and No. 26-cv-09554. Both are signed by the same lawyer, Steven Lieberman of Rothwell Figg, and follow the same five-count structure.

The suits matter beyond one browser company because of what they attack. Most AI copyright litigation has centered on whether training a model on news is lawful. These complaints go after the retrieval layer: an independent web index, sold by the request, that AI products use to ground answers. Anyone building on a third-party search API should read them. As of the latest coverage, Brave had not publicly responded, and everything below about its conduct is an allegation that no court has tested.

What the publishers say they got out of the API

The complaints describe an index that Brave itself puts at more than 30 billion pages, reached through four API endpoints: News Search, Web Search, Answers (which the complaints say is the service behind the Ask Brave product) and LLM Context, which returns extracted chunks of pages so that developers do not have to scrape them.

The publishers’ test results are the most concrete part of the filings. According to PPC Land’s reading of the complaints:

• LLM Context returned 1,291 of the 1,295 words of a Times article about a robot park, verbatim.
• The same endpoint returned the entire text of an Orlando Sentinel article on a Florida amendment, while News Search returned 250 words of it.
• News Search returned nearly 350 words of a Daily News article and more than 250 words of a Chicago Tribune article.

Ask Brave, the consumer answer product, allegedly produced more than 1,200 words of a Tribune article on data center tax breaks and nearly 600 words of a Daily News article on Central Park horse carriages. In one Times test it reportedly delivered about 87% of an article while telling the user the remainder sat behind the paywall.

Those numbers do most of the legal work. A model that summarizes a story in two sentences raises a hard question about transformation. An endpoint that returns 99.7% of an article’s words to a paying customer raises a much simpler one.

Five counts, and why two of them are unusual

Each complaint pleads direct, contributory and vicarious copyright infringement, circumvention under Section 1201 of the Digital Millennium Copyright Act, and unjust enrichment under New York law. Neither names a dollar figure. Both ask for statutory and actual damages, a permanent injunction and a jury trial.

The Section 1201 count targets a browser feature rather than the index. Brave’s Shields filter lists include an entry the complaints describe as a paywall blocker, labeled with the name of the Bypass Paywalls Clean filters. Fair use, the position Brave has taken in earlier litigation, does not map neatly onto a circumvention claim, because the claim concerns getting past the access control and not what is done with the text afterward. Damages are also counted differently: Section 1203 allows $200 to $2,500 per act of circumvention, and the complaints treat each act as a separate violation. The complaints do not say whether the setting is on by default, and that is likely to matter.

The unjust enrichment count is aimed at Brave’s business model. Brave blocks ads by default and sells its own advertising, and the publishers want the resulting profits disgorged. The complaints state that this claim does not depend on any copying. Our read is that the framing is deliberate, since state-law claims that merely restate copyright infringement tend to be preempted by federal law. It is also the count with the widest possible reach: if blocking a publisher’s ads while selling your own is unjust enrichment, the theory is not limited to AI.

The undisclosed crawler problem

Both complaints stress that Brave has never named its crawler. Brave’s own help page explains why: it uses no distinct user agent, because some sites allow only Google to crawl, and it says its bot will not crawl pages that Googlebot cannot reach.

For publishers that creates a bind. A robots.txt rule cannot single Brave out, so the only way to exclude it through that mechanism is to exclude Google too, which few news businesses could afford. The Times’s letters to chief executive Brendan Eich, dated August 11, 2025 and April 13, 2026, reportedly went unanswered, and the second one warned of willful infringement and enhanced statutory damages of up to $150,000 per work. If proven, that notice would help the Times argue willfulness. The Daily News complaint describes no such letters and rests its willfulness argument on the paywall bypass and the crawler.

Anthropic, xAI and the other customers in the filings

Brave is the only defendant, but the complaints name Anthropic, Perplexity, Mistral, xAI and Cohere as customers of its index. They quote a November 2023 Brave post saying Anthropic used the Search API to train Claude 2, and cite Meta’s Llama 3 paper for using the API to answer questions about recent events. The publishers’ lawyer made the implied warning explicit, telling TheWrap: “Buying stolen goods secondhand doesn’t make them legitimate.”

That is the piece for AI builders to weigh. A licensed-looking API contract does not settle whether the content behind it was cleared, and the vicarious and contributory theories here depend on what downstream customers do with the text. Teams that use search APIs for grounding have practical questions to ask now: what indemnity the provider offers, whether full-page extraction is really needed when snippets would do, and whether they log what was retrieved and displayed to users. The Times and the three Tribune-group papers are already plaintiffs against OpenAI and Microsoft, as other regional publishers have since done, and the Times sued Perplexity in December 2025. The target list is moving down the stack.

What to watch

Brave’s first filing will show whether it leads with fair use or attacks the tests themselves, since the outputs were generated by the plaintiffs’ own queries. The related dispute with News Corp is a preview: Brave sued first in March 2025 seeking a declaration that its indexing is fair use, and News Corp countersued that July. A ruling there could land before anything happens in these cases.

The other marker is any request for early relief. The complaints seek a permanent injunction, which would come only at the end of the case. A preliminary order touching LLM Context or the Answers endpoint would affect every product built on them much sooner. Meanwhile, the undisclosed crawler argument puts pressure on an industry norm, and publishers will want a ruling that treats it as evidence of bad faith.