The New York Times Co. v. Brave Software and Daily News, LP v. Brave Software

A retrieval case, not mainly a training case

Most AI copyright suits are about what went into a model during training. These two are mostly about what an AI product fetches and repeats when someone asks it a question. The complaints say Brave copies pages into the Brave Index with a crawler whose bot name, they allege, Brave has never disclosed, leaving site owners no way to block it (NYT ¶¶ 2, 77). That index feeds Ask Brave, which writes answers grounded in retrieved articles, and the Brave Search API, which Brave sells to other companies (NYT ¶¶ 4, 53-68).

Training appears as a secondary theory. The complaints say Brave sells copies of its index to AI developers for training or for grounding their answers, and quote Brave’s own marketing describing those as the API’s most common uses (NYT ¶¶ 5, 61-62). The contributory-infringement count rests on those customers’ use (NYT ¶ 135; Daily News ¶ 110). Neither complaint alleges that Brave trained a model of its own.

The API customers are named but not sued

The NYT complaint names Anthropic, Perplexity, Mistral, xAI and Cohere as API customers (¶ 1). It also names You.com, says Meta trained Llama 3 to use the API, and cites a Snowflake integration, relying on public reports (¶ 62 and notes). None of them is a defendant. The complaints aim at the supplier of the index rather than its users.

Five counts

What the complaints say the plaintiffs tested

Both complaints set out examples the papers say they ran themselves. For Ask Brave, the Times says that when given only the address of a paywalled article, the feature summarized it, offered the full text, and then produced more than 400 words of it verbatim (NYT ¶¶ 79-80). Other examples describe outputs reproducing most or nearly all of individual articles (NYT ¶¶ 82-86). For the API, the Times says one LLM Context response returned 1,291 of an article’s 1,295 words verbatim (NYT ¶ 97), and that for another article the LLM Context endpoint returned nearly half and the Web Search and News Search endpoints more than a third each (NYT ¶¶ 94-96). The Daily News complaint reports similar results for each of its three papers, with one side-by-side comparison attached as Exhibit D (Daily News ¶¶ 54-70).

These are the plaintiffs’ own tests, described in their pleadings. Brave has not yet answered them, and nothing in the complaints shows how often ordinary users receive such outputs.

The paywall and advertising claims

The § 1201 count is about the Brave Browser rather than the search products. The complaints say its paywall blocker lets users load subscriber-only articles without signing in, and that it removes subscription prompts (NYT ¶¶ 70-74, 99-104; Daily News ¶¶ 72-74). The unjust-enrichment count says Brave strips the papers’ ads and serves its own in their place (NYT ¶¶ 98-105, 155-162).

Notice and willfulness

The Times says it sent cease-and-desist letters to Brave’s chief executive in August 2025 and April 2026 and says Brave did not stop (NYT ¶¶ 49-50). It pleads willfulness on those letters, the paywall bypass and the unnamed crawler (NYT ¶ 125). The Daily News complaint relies on the paywall bypass and the crawler, and the copy reviewed does not describe any letters (Daily News ¶ 100).

Where these sit

The Times and the Daily News papers are also plaintiffs in the news cases against OpenAI and Microsoft (see In re OpenAI). ChatGPT Is Eating the World says these two suits brought its count of copyright suits against AI companies in U.S. courts to 160, and notes that Brave is itself a plaintiff against News Corp in a separate California case.