New York Times and three other newspapers sue Brave over search index sold to AI companies

The products targeted by the complaints:

• Brave Search API. Four endpoints: News Search, Web Search, LLM Context (page chunks shaped for LLM input) and Answers. Pricing is about $5 per 1,000 search requests, or $4 per 1,000 Answers requests plus $5 per million tokens (DEV Community).
• Ask Brave. A combined search and AI chat product launched September 29, 2025, which uses retrieval-augmented generation (RAG) over the index (PPC Land).
• Browser features. A "Bypass Paywalls Clean Filters" adblock list, and Brave Ads, which replace blocked publisher ads (PPC Land).

• Crawler identity. Brave's own help page says its crawler deliberately has no distinct user agent, to avoid being shut out by sites that allow only Google. It also says it won't crawl anything Googlebot is barred from. For delisting, robots.txt doesn't work; sites must use a noindex directive (Brave Search help). The practical result is that publishers can't allow Google while blocking Brave. That is the factual core of the "unblockable" allegation.
• Retrieval, not training. The case centres on RAG outputs and API redistribution. The LLM Context endpoint returns extracted page text, so liability could attach at the moment of retrieval, not only during model training.
• Plaintiff tests. The complaint says Ask Brave reproduced nearly 700 words of one Times article verbatim and about 87% of another (PPC Land). The test prompts and method are not public.

• Index scale and crawler. The 30-billion-page index and the hidden crawler are alleged in the complaint (TechCrunch). The crawler behaviour matches Brave's own documentation (Brave).
• Named AI customers. Anthropic, Perplexity, Mistral, xAI and Cohere are named in the complaint. Coverage also lists Meta and You.com (PPC Land). None of these companies has confirmed the claims in the reviewed coverage.
• Verbatim reproduction. This comes from the plaintiffs' own testing. No independent reproduction exists.
• Brave's figures. Revenue above $100 million in 2025 and about 130 million monthly users come from secondary reporting (PPC Land). Total funding of about $250 million is a private-market-tracker estimate (TechCrunch).

• The Times is already suing OpenAI and Microsoft; summary judgment motions in that case were filed September 4, 2026. The Times and the Chicago Tribune sued Perplexity in December 2025, and CNN sued Perplexity in May 2026 (PPC Land).
• The new suits move one layer down the supply chain: from AI products to the search index that feeds them (Search Engine Journal).

• All allegations are unadjudicated. Brave had not responded publicly by Monday evening (The Next Web; Dataconomy).
• In the earlier News Corp dispute, Brave argued that indexing, snippets and summaries are fair use and standard search practice (WDEZ/Reuters). That defence is likely to be contested again.
• Brave's documentation implies it honours Googlebot restrictions. This partly conflicts with the plaintiffs' claim that publishers have no means to block Brave at all.
• The unjust-enrichment count over ad replacement doesn't depend on any copying (PPC Land).

• AI builders on third-party search APIs now carry supply-chain risk. Practical steps:
• Log which provider supplied each citation.
• Check what your API plan allows you to store.
• Put the search provider behind an abstraction so it can be swapped (DEV Community).
• Publishers gain a path to pressure intermediaries, not just model labs.
• Search and crawler operators face a tradeoff: an undisclosed user agent may now be a legal liability, not just a competitive tactic.
• Requested remedies: injunctions, statutory damages and disgorgement of profits.