On Monday 2 November 2026, The New York Times sued Brave Software in the US District Court for the Southern District of New York (reported, court filing, case 26-cv-9549). The same day, the Daily News, the Chicago Tribune and the Orlando Sentinel filed a separate complaint against Brave in the same court (reported, court filing, case 1:26-cv-09554).
Most coverage has framed this as another AI copyright case. It is more specific than that, and more useful. The complaints describe, step by step, a working supply chain for the real-time reading of publisher content by AI systems. Content is crawled, indexed, chunked for language models and sold to AI companies at query time. According to the plaintiffs, every link in that chain earns money except the first one.
What does the lawsuit actually allege?
The Times describes Brave as "the supplier, dealer, and distributor of stolen copyrighted content for its own users and many of the most prominent GenAI companies" (named-party figure, NYT complaint). The allegations fall into four parts.
First, the crawl. Brave is alleged to have crawled "over 30 billion webpages" for its search index without identifying its bot, and without giving site owners a way to block it (named-party figure, both complaints).
Second, the answer engine. Brave launched a generative AI feature called Summarizer in Brave Search in 2023 and renamed it Ask Brave in 2025. The complaints say it "often incorporates verbatim, third-party content," using retrieval-augmented generation that Brave calls "AI Grounding" (named-party figure, both complaints).
Third, the API. The complaints allege that "other AI companies pay Brave to provide them with real-time online content from the Brave Index," through endpoints including Web Search, News Search, LLM Context and Answers. Brave's own description of LLM Context, quoted in both filings, is that it "returns extracted page chunks and source metadata in a shape and length that satisfies the needs of an LLM without need for further scraping."
Fourth, the browser. The complaints point to a "Paywall blocker" setting that Brave promoted as letting users "Bypass Paywalls," and to Brave blocking publishers' ads and showing its own (named-party figure, both complaints).
The Times complaint names Anthropic, Perplexity, Mistral, xAI and Cohere as customers of Brave's API (named-party figure, NYT complaint). Those companies are not defendants, and the claims about them are allegations.
How much of an article does an AI answer take?
The plaintiffs put numbers on it. In one example cited from the Times complaint, the LLM Context endpoint returned 1,291 of an article's 1,295 words (reported, CyberInsider, citing the NYT complaint). The other three publishers say Ask Brave generated more than 1,200 words from a Chicago Tribune article on data-centre tax breaks, and nearly 600 words from a Daily News story on Central Park horse carriages (reported, TheWrap, citing the complaint).
That is not a snippet. It is the read itself, delivered to a machine.
The money side is stated just as bluntly: "While Brave has made more than $100 million by copying and reselling publishers' work without consent or permission, publishers have received nothing" (named-party figure, both complaints). Their complaint cites third-party reports that Brave's 2025 revenue passed $100 million, up from about $52 million in 2024 (named-party figure, citing third-party reports). It also cites close to 130 million monthly users, over 50 million daily users and more than 50 million search queries a day as of 30 September 2026 (named-party figure, complaint).
Brave had not publicly responded as of 4 November (reported, CyberInsider). The Times says it wrote to chief executive Brendan Eich on 11 August 2025, and again through outside counsel on 13 April 2026, and that Brave did not reply to either letter (named-party figure, NYT complaint).
Why is this case different from the Times's earlier AI suits?
Most publisher litigation so far has centred on training: whether copying archives to build a model is fair use. The Brave complaints are mostly about something else, the live read. Grounding APIs fetch current content at the moment a user asks a question, and pass it to a model to compose the answer.
Definition: an AI grounding API is a service that retrieves current web content (pages, chunks or extracted passages) and supplies it to a language model at query time, so the model's answer reflects up-to-date sources. It powers the real-time reading done by Live Search Agents, as distinct from the bulk collection done by training crawlers.
That distinction matters commercially. A training copy happens once. A grounding read happens every time someone asks a question, which makes it recurring, intent-driven and countable. The Brave complaints are, in effect, an argument about who owns the revenue from that recurring read.
The case also extends a longer fight. TheWrap reports that Brave had earlier sued News Corp seeking a ruling that its search practices were lawful, and that News Corp countersued in July 2026 (reported, TheWrap).
And the market is moving in parallel. On 4 November, Adweek reported that Anthropic, one of the companies the Times names as a Brave API customer, had signed its first content licensing deal with a news publisher: the Associated Press, supplying AP-attributed election-night data in Claude (reported, Adweek, citing two people familiar with the matter). Neither company commented on the record, and no terms were disclosed. If accurate, it suggests that AI firms alleged to buy real-time content through intermediaries will also pay a publisher directly when the content is time-critical.
What does it mean for publishers?
Three things.
The read has a market price. AI companies are, according to the complaints, already paying for real-time access to news content. The question is no longer whether a machine read is worth money. It is who collects.
Blocking is not a plan. The complaints allege a crawler that did not identify itself and offered no opt-out. A robots.txt strategy assumes the reader announces itself; the most valuable readers may not. And even a perfect block only sends the AI user to whichever source stayed open.
Visibility is the precondition for pricing. Publishers cannot charge for reads they cannot see. Live Search Agents typically do not run JavaScript, so no analytics fire and no ad tags load. The Brave case was built on manual tests showing what came out the other end. Most publishers have no equivalent view of what goes in.
What does it mean for advertisers and GEO?
For brands, the grounding layer is where AI answers are assembled. Whatever content a grounding API can reach, at the moment a user asks, shapes which products and brands appear in the response. If courts or deals push that layer towards licensed, permissioned content, the sources an answer draws on become more defined, and more measurable.
That makes generative engine optimisation (GEO, earning organic presence in AI answers) partly a question of supply: which publishers' pages are in the grounding set. It also strengthens the case for paid, contextual brand presence at the point where an agent reads a trusted publisher page. Display, paid search and social do not appear inside an AI answer. The read is where the answer is decided.
the blankspace read: Strip away the copyright language and the Brave complaints are a price list. They describe AI companies paying, per request, for real-time reads of journalism. That is the market blankspace has been describing: publishers are read by AI at scale and paid almost nothing. What the lawsuit adds is evidence that the payment already exists; it is simply being collected one step down the pipe, by an intermediary. Litigation can contest that, slowly, one defendant at a time. The faster fix is for publishers to see every agent read and charge for it at the source. Monetise the read, not the referral, before someone else monetises it for you.
Key figures
| Figure | Value | Provenance |
|---|---|---|
| Filing date and court | 2 November 2026, SDNY | Reported (court filings) |
| Case numbers | 26-cv-9549 (NYT); 1:26-cv-09554 (Daily News, Chicago Tribune, Orlando Sentinel) | Reported (court filings) |
| Copyrighted works owned by the Times | Over 10 million | Named-party figure (NYT complaint) |
| Webpages Brave allegedly crawled | Over 30 billion | Named-party figure (complaints) |
| Words returned from one Times article via LLM Context | 1,291 of 1,295 | Reported (CyberInsider, citing NYT complaint) |
| Words generated from one Chicago Tribune article by Ask Brave | More than 1,200 | Reported (TheWrap, citing complaint) |
| Brave's alleged earnings from publishers' work | More than $100 million | Named-party figure (complaints) |
| Brave revenue, 2024 to 2025 | About $52m to over $100m | Named-party figure (complaint, citing third-party reports) |
| Brave users and queries (30 Sept 2026) | ~130m monthly users; 50m+ daily users; 50m+ queries a day | Named-party figure (complaint) |
| Unanswered letters from the Times | 11 August 2025; 13 April 2026 | Named-party figure (NYT complaint) |
| Anthropic and AP deal terms | Not disclosed | Reported (Adweek, anonymous sources) |
Primary source, opened and verified: The New York Times Company v. Brave Software, Inc., complaint, SDNY 26-cv-9549, filed 2 November 2026 (CourtListener). Companion complaint: Daily News, Chicago Tribune and Orlando Sentinel v. Brave Software, SDNY 1:26-cv-09554 (CourtListener). Supporting reporting: TheWrap (2 November 2026), CyberInsider (4 November 2026), Adweek on Anthropic and the AP (4 November 2026).