Publishers Suing AI Companies: The Brave and Anthropic Cases Explained

Key Takeaways

• On 2 November 2026, The New York Times and three other US newspapers filed two lawsuits against Brave Software in federal court in Manhattan, accusing it of copying their journalism into a search index it sells to AI companies.
• The US cases name Anthropic, Perplexity, Mistral, xAI and Cohere as customers of Brave’s search API, but none of those companies is a defendant in either case.
• In Canada, Postmedia, Torstar, The Globe and Mail and The Canadian Press filed a copyright claim against Anthropic in an Ontario court on 6 October 2026, alleging Claude was built on scraped journalism.
• Both sets of cases turn on the same question: whether copying published work to build AI systems and AI answers is infringement, or whether it is fair use.
• Nothing has been decided yet. These are allegations in complaints, not findings by any court, and the cases could take years.
• The outcome matters far beyond the courtroom, because it will shape whether AI companies pay to license news, whether publishers can block crawlers, and what AI search answers are allowed to show.

Two Lawsuits in One Day

On Monday 2 November 2026, The New York Times Company sued Brave Software in the United States District Court for the Southern District of New York. Hours later, the publishers of the New York Daily News, the Chicago Tribune and the Orlando Sentinel filed a second, closely matching lawsuit in the same court. Both complaints name a single defendant, Brave Software, the San Francisco company behind the Brave browser and Brave Search. Both demand a jury trial. Both were signed by the same lawyer, Steven Lieberman of the Washington firm Rothwell, Figg, Ernst and Manbeck.

The case numbers tell the story of a coordinated filing. The Times case is numbered 26-cv-9549. The three-paper case is 26-cv-9554. The Times complaint runs to 72 pages. The complaint from the three papers runs to 69 pages. The three papers are all owned by Tribune Publishing, which is controlled by Alden Global Capital, and their complaint was filed by Daily News, LP, Chicago Tribune Company, LLC and Orlando Sentinel Communications Company, LLC.

The timing is worth getting right, because the headlines made the cases sound newer than they are. The Brave lawsuits were filed on 2 November, twelve days before this article was written. The Canadian case against Anthropic is older still: it was filed on 6 October and first reported publicly on 19 October. Both are recent, but neither is breaking news, and the distinction matters because the first responses, or the lack of them, are already part of the story.

What the publishers are suing over is not a chatbot. Brave does not make a large language model. It makes a browser and a search engine, and the complaints attack the layer underneath AI answers: the search index. That is what makes these cases different from the lawsuits publishers have already filed against OpenAI, Microsoft and Perplexity. The target here is the supplier, not the model builder.

What the Newspapers Say Brave Did

The complaints make four main claims, and they are worth separating because each one points at a different Brave product.

The first claim is about the index itself. The publishers say Brave built a search index of more than 30 billion pages by crawling and copying websites, including their articles, without permission. The crucial detail is that Brave’s crawler does not identify itself under a distinct name, so publishers cannot block it by name in the way they can block a crawler that announces itself. The complaints say this was deliberate: Brave built its index in a way that publishers could not keep it out.

The second claim is about the Brave Search API. The publishers say Brave sells access to its index through a paid API, and that AI companies are among the buyers. The complaints name five of them: Anthropic, Perplexity, Mistral, xAI and Cohere. The Times complaint goes further, quoting Brave’s own marketing to argue that training AI models and grounding AI answers in live web content are among the most common uses of the API. One strand of the complaint is particularly concrete. The Times says it tested the API’s LLM Context endpoint, which returns page content for AI tools, and found that it returned 1,291 of the 1,295 words in a Times article. That is the kind of test result that turns an abstract claim about copying into something a judge can picture.

The third claim is about Ask Brave, the AI answer feature inside Brave Search. The publishers say Ask Brave reproduces passages from their articles word for word and produces detailed summaries of paywalled stories, giving readers the substance of the journalism without ever visiting the newspaper’s website. Brave launched an AI summary feature called Summarizer in 2023 and later updated and renamed it Ask Brave in 2025. The complaints treat the answers as a substitute for the original articles, not as a pointer to them.

The fourth claim is about the browser. The complaints say the Brave browser has a setting called Paywall blocker, which Brave itself describes as a way to bypass paywalls, and that the browser strips publishers’ advertising while showing Brave’s own. This is the claim that turns a copyright case into a business-model case: the publishers are not only saying their words were copied, they are saying the money that should have followed those words was taken too.

The legal counts follow from those claims. Each complaint pleads five counts: direct copyright infringement, contributory copyright infringement, vicarious copyright infringement, circumvention of technological protection measures under the Digital Millennium Copyright Act, and unjust enrichment under New York law. The publishers are seeking damages, disgorgement of advertising revenue and a permanent injunction. Neither complaint names a dollar figure for damages.

The language in the complaints is deliberately sharp. The Times calls Brave the supplier, dealer and distributor of stolen copyrighted content, and describes it as acting like a burglar towards publishers and an AI fence towards the companies it supplies. Those are the plaintiffs’ words. No judge has adopted them, and they should be read as advocacy, not as findings.

The Letters Before the Lawsuits

The complaints say the lawsuits did not come out of nowhere. According to the Times, it sent cease-and-desist letters to Brave chief executive Brendan Eich on 11 August 2025 and again on 13 April 2026, and Brave ignored both. A Times spokesperson, Graham James, told The Verge that Brave has flouted established copyright law and ignored repeated requests by The Times to stop its infringing conduct.

If that account is right, it matters for more than the narrative. In copyright cases, evidence that a defendant was warned and carried on anyway can support a claim that the infringement was wilful, and wilfulness can increase the damages a court awards. That is a long way from being decided, and Brave has not yet answered the claims in court, but the letters are likely to be a central part of the argument about intent.

The complaints also lean heavily on Brave’s own words. They quote a November 2023 announcement in which Brave said Anthropic had used the Brave Search API to train Claude 2 on high-quality data. They quote Brave’s documentation describing the API’s LLM Context endpoint as giving AI tools page content without need for further scraping. And they quote Brave’s help page on its search crawler, which says the crawler does not advertise a differentiated user agent because Brave must avoid discrimination from websites that allow only Google to crawl them. That last point is the heart of the dispute about the unnamed crawler: Brave says it hides the crawler’s identity because sites that admit only Google would shut it out; the publishers say hiding the crawler’s identity is exactly what stopped them from blocking it.

What Brave Has Said So Far

As of the reporting that followed the filings, Brave had not responded. TechCrunch reported that Brave did not immediately respond to a request for comment. The Verge reported the same. As of 3 November, Brave’s blog showed no post about either suit. That silence is not an admission of anything, and it is not unusual in the first days of a lawsuit, when a company is taking legal advice before saying anything in public. But it does mean that everything in the public record about the Brave cases so far comes from the publishers’ side.

One company named in the complaints did push back publicly, and it was not Brave. Perplexity, which the Times has sued separately, sent a dismissive comment to The Verge, saying the Times is overpaying its attorneys and underpaying its journalists. That is a public-relations jab, not a legal argument, and it says nothing about the merits of the Brave cases. Anthropic, Mistral, xAI and Cohere did not immediately reply to requests for comment, according to the reporting at the time. It is worth repeating that none of the five is a defendant in either Brave case. Being named as a customer in a complaint is not the same as being accused of wrongdoing, and the complaints do not accuse them of any.

Brave is not a stranger to this fight. The company previously filed a declaratory judgment action against News Corp in federal court in Oakland, asking the court to declare that its web indexing is fair use, and News Corp countersued in July 2026. That earlier case shows Brave has been prepared to litigate the legality of its index rather than negotiate over it, and it suggests the company may take a similar position here: that building a search index from publicly available web pages is lawful, and that the publishers’ real complaint is with a business model they do not like. That is a prediction, not a fact, and Brave has not yet said so in these cases.

The Canadian Case Against Anthropic

The second half of this story is a different lawsuit, in a different country, against a different kind of defendant. On 6 October 2026, a group of Canadian news publishers filed a statement of claim in an Ontario court against Anthropic, the company behind Claude, and its Canadian subsidiary. The claim became public when the news site Cartt.ca first reported it, and CBC News and The Globe and Mail then reported it on 19 October, carrying a Canadian Press report.

The plaintiffs are Postmedia, Torstar, The Globe and Mail and The Canadian Press. Together they publish a large share of Canada’s daily journalism: Postmedia is the country’s largest newspaper chain and publishes the National Post and many big-city dailies, Torstar owns the Toronto Star and other Ontario titles, and The Canadian Press is the national news agency that supplies stories to outlets across the country. Some reports of the filing also name affiliated titles, including the Toronto Star, Metroland Media and PNI Maritimes, as part of the plaintiff group.

The claim alleges that Anthropic infringed the publishers’ copyright by systematically harvesting and misappropriating millions of works to train and operate its commercial AI models, including Claude, without permission, licensing agreements or financial compensation. The publishers say Anthropic used automated web crawlers, including one called ClaudeBot, to scrape their content for commercial purposes, breached their websites’ terms of service, circumvented paywalls and ignored restrictions that limited use to personal purposes. The claim is brought under Canada’s Copyright Act.

The scale of the alleged copying, as described in the claim, is large. Reporting on the filing says The Canadian Press owns the largest number of works allegedly scraped, at 4.2 million, followed by Postmedia and PNI Maritimes at 1.75 million. One report describes the publishers as seeking potentially billions of dollars in damages, while other reports note that no dollar figure was given in the published reporting of the claim. That discrepancy is worth flagging rather than smoothing over: the size of the claim has not been confirmed in a single consistent figure, and the claim itself has not been made public in full.

The claim also points to Anthropic’s size and success, noting that as of May the company had raised approximately 125 billion dollars in capital and was valued in excess of 965 billion dollars, with more than a million users signing up for Claude per day. Those figures come from the publishers’ own claim, as reported, and they are part of the argument that Anthropic profited from the alleged copying on an enormous scale. They have not been tested in court.

Anthropic had not responded to requests for comment when the case was reported, and it had not yet filed a statement of defence. Under Ontario procedural rules, foreign defendants served outside the country have up to 60 days to respond, so the absence of a defence in the first weeks says nothing about the company’s position. As with the Brave cases, none of the allegations has been proven in court.

The same group of publishers is already suing OpenAI. They launched a joint lawsuit against OpenAI in November 2024, together with CBC and Radio-Canada, and that case is still unresolved. The Anthropic case is therefore the second front in a Canadian campaign, not the first, and it suggests the publishers have decided that litigation, rather than negotiation alone, is how they intend to deal with AI companies that train on their work.

The Wider Copyright War

These two cases do not exist in isolation. They are the latest fronts in a copyright war that now stretches across the United States, Canada and Europe, and across books, music, images and news. Understanding the Brave and Anthropic cases means seeing where they sit in that larger fight.

The New York Times is already suing OpenAI and Microsoft, in a case filed in December 2023 in which summary judgment motions were filed on 4 September 2026. The Times has also sued Perplexity. The Canadian publishers, as noted, are already suing OpenAI. Anthropic itself reached a 1.5 billion dollar settlement with book authors and publishers in the Bartz case, which alleged copyright infringement involving literary works used to train its models; a district court approved that settlement in July 2026. Some authors and publishers opted out of that settlement and have brought individual lawsuits instead.

Anthropic also faces claims from music publishers. Concord Music Group and other music publishers sued Anthropic in October 2023, alleging it copied and distributed copyrighted song lyrics, and further complaints from publishers including Universal Music Publishing Group, Sony Music Publishing and Warner Chappell were filed between January and October 2026. Those cases are at a preliminary stage. Reddit sued Anthropic in June 2025 over its use of Reddit data; in September 2026 a court narrowed that case to claims involving breach of contract, interference with contract and unfair competition. In Canada, Anthropic faces four copyright actions in total, including two proposed class actions on behalf of authors filed in 2025 and a lawsuit from the Quebec news publisher La Presse filed in July 2026.

Against that background, one recent development stands out because it points in the opposite direction from litigation. Adweek reported on 4 November 2026 that Anthropic had signed its first content licensing deal with a digital news publisher, the Associated Press, struck just before the US midterm elections. According to the report, which cited two people familiar with the matter, Claude showed up-to-date AP election information with attribution on election night. Neither Anthropic nor the AP commented on the record, and the financial terms and duration were not disclosed. The deal is therefore reported, not confirmed, and Adweek itself described it as a limited test rather than a sign that Anthropic has fully changed its approach to licensing. But if the report is right, it is the first time Anthropic has paid a digital news publisher for content, and it sits awkwardly alongside the company’s position in the Canadian case that its use of publishers’ work is lawful without a licence.

The European picture adds another layer. Our report on the EU AI copyright consultation that closed on 3 November explains how European regulators are trying to write rules for exactly this problem: whether AI companies must respect opt-outs, disclose what they trained on, and pay for what they use. The consultation closed the day after the Brave lawsuits were filed, which is a coincidence, but it is a telling one. Courts and regulators on both sides of the Atlantic are being asked to answer the same question at the same time, and they may not give the same answer.

What Is Actually at Stake

Strip away the legal language and the stakes are straightforward. Publishers make money in two ways: readers pay for subscriptions, and advertisers pay to reach those readers. AI search answers threaten both. If an AI answer gives a reader the substance of a story, the reader has less reason to click through, subscribe or even see the advertising that funds the reporting. That is the publishers’ core complaint, and it is an economic one as much as a legal one.

The AI companies’ answer, in general terms, is that training on publicly available text and summarising it is fair use, that search indexes have always copied the web in order to point people to it, and that AI answers drive at least some traffic back to sources. That is the argument Brave is likely to make about its index, and it is the argument Anthropic has made about training in other cases. Whether a court accepts it will depend on the specific facts: how much was copied, whether the copying substitutes for the original, whether the use was transformative, and what effect it had on the market for the original work. Those are the classic fair-use questions, and they do not have a single answer that covers every case.

Three more specific issues will shape how these cases play out.

The first is the crawler problem. The web has a long-standing system for controlling crawlers: the robots.txt file, in which a site tells automated visitors what they may and may not fetch. But that system only works if crawlers identify themselves. The Brave complaints allege that Brave’s crawler did not, which meant publishers could not block it even if they wanted to. If a court finds that the unnamed crawler was a deliberate evasion of publishers’ controls, that could matter a great deal, and not only for Brave. Our earlier look at the OpenAI cross-company web crawler signal shows the same question coming up with other companies’ crawlers. The principle at stake is simple: a publisher that cannot identify a crawler cannot exercise any choice about it, and a system built on choice fails if the choice cannot be exercised.

The second is the difference between training and retrieval. Training a model on a text and retrieving a text to answer a question are different acts, and the law may treat them differently. The Canadian case is mainly about training: the claim is that Claude was built on scraped journalism. The Brave cases are mainly about retrieval and redistribution: the claim is that Brave’s index and API deliver publishers’ articles, in full or in substance, to AI systems and AI answers. A court could find one lawful and the other not, or find both lawful, or both unlawful. The distinction is one reason these cases are worth following separately rather than lumping together as one fight about AI and copyright.

The third is licensing. The AP deal, if confirmed, shows that at least one AI company is willing to pay at least one publisher, at least for election data. OpenAI has signed more than a dozen publisher licensing deals, according to Adweek’s reporting. If licensing becomes the norm, the Brave and Anthropic cases may end up as negotiating leverage as much as litigation: the threat of a large damages award pushes companies towards paying, and the existence of paid deals weakens the argument that licensing is impossible. But licensing also raises its own questions. A market in which only the largest publishers can strike deals leaves smaller outlets with nothing, and a market in which AI companies pay for some content and scrape the rest is not a market at all. Those questions will not be answered by a single verdict.

What Happens Next

In the Brave cases, the next step is Brave’s answer to the complaints. The company will have to respond to the allegations in court, and its response will reveal its defence: whether it argues fair use, whether it disputes the facts about the crawler and the API, and whether it challenges the publishers’ claims about the paywall setting and the advertising. No hearing date has been reported, and no ruling on the merits should be expected soon. Cases of this kind typically take years, and they often settle before a verdict.

In the Canadian case, the next step is Anthropic’s statement of defence, which is due within the period allowed for a foreign defendant served outside Canada. The case will then move through the Ontario courts, and it may run alongside the publishers’ earlier case against OpenAI, which has been going since November 2024 without resolution. The pace of Canadian litigation is not fast, and readers should not expect a quick answer.

There are also developments to watch outside the courtroom. If Anthropic signs further publisher deals after the reported AP agreement, that will suggest licensing is becoming a real strategy rather than a one-off test. If Brave responds publicly, or changes how its crawler identifies itself, that will say something about how seriously it takes the publishers’ central claim. And if the Times’ case against OpenAI and Microsoft produces a ruling on summary judgment, that ruling will shape the arguments in every other publisher case, including these two. Our coverage of Anthropic’s oversubscribed IPO is also relevant here, because a company heading for the public markets has more reason than most to settle its legal exposure, and Anthropic’s IPO filing itself describes the copyright suits in the United States and Canada.

For readers, the practical consequences are closer than the court dates. If the publishers win, AI search answers may start showing less news content, or linking more prominently to sources, or carrying licensing labels. If the AI companies win, the current shape of AI answers is likely to continue, and publishers will have to find other ways to get paid for work that AI systems summarise. Either way, the way you get news from an AI assistant in a year’s time may look different from the way you get it today, and these two cases are part of the reason why.

What This Means for AI Users

There is a temptation to treat publisher lawsuits as a fight between large companies that has nothing to do with ordinary users. That misses the point. The outcome will decide what AI assistants are allowed to tell you about the news, and how they are allowed to tell it.

If AI companies must license news content, they may pass the cost on, or they may limit news answers to licensed sources, which would narrow the range of what they can answer. If they do not have to license it, the answers may stay broad and free, but the journalism those answers depend on may get harder to fund. Neither outcome is obviously good or bad for users, and the trade-off is real: free, comprehensive AI answers and well-funded independent journalism are both things worth having, and at the moment they are in tension.

The honest position is that nobody knows how this ends. The law on AI training is unsettled, the facts in these cases are disputed, and the courts have not ruled. What is certain is that the argument has moved beyond blog posts and policy papers into courtrooms, and that the companies involved are now spending real money on it. The Brave cases and the Canadian Anthropic case are two more steps in that direction, and they deserve to be read carefully rather than cheered or dismissed.

What We Could Not Verify

• Whether Brave has filed any response to the two lawsuits since the initial reporting on 2 and 3 November 2026. At the time of the reporting we relied on, Brave had not responded publicly or in court.
• The exact dollar figure the Canadian publishers are seeking. One report describes the claim as seeking potentially billions in damages; other reports say no dollar figure was given in the published reporting. We have not seen the statement of claim itself.
• Whether the reported Anthropic and Associated Press licensing deal has been confirmed by either company. Adweek reported it on 4 November 2026 citing two people familiar with the matter; neither company commented on the record and no terms were disclosed.
• The precise contents of the complaints beyond what the reporting describes. We relied on reporting of the filings, not on the filed documents themselves, and some details, such as the full list of affiliated plaintiff titles in the Canadian case, differ between reports.
• Whether Brave’s crawler identification practice has changed since the lawsuits were filed, and whether the company has made any public statement after 3 November 2026.

Sources

• TechCrunch, “New York Times and other outlets sue web browser Brave over AI scraping,” 2 November 2026, as reported by PPC Land, “New York Times and three papers sue Brave over AI answers and ad blocking,” 3 November 2026: https://ppc.land/new-york-times-and-three-papers-sue-brave-over-ai-answers-and-ad-blocking/
• The Verge, “The New York Times sues Brave web browser for AI scraping and bypassing paywalls,” 2 November 2026
• Copyright Now, “Daily News v. Brave Software (S.D.N.Y. 1:26-cv-09554): newspaper AI search and paywall bypass complaint,” 2 November 2026: https://copyrightnow.com/news/daily-news-chicago-tribune-orlando-sentinel-v-brave-software-copyright-dmca-complaint-sdny/
• Inside Telecom, “Generative AI in Journalism Turns on Paywall Ads,” 5 November 2026: https://insidetelecom.com/generative-ai-in-journalism-turns-on-paywall-ads/
• CBC News and The Canadian Press, via The Globe and Mail, reporting the Canadian publishers’ claim against Anthropic, 19 October 2026
• Broadcast Dialogue, “Canadian news publishers sue Anthropic over AI copyright infringement,” 14 October 2026: https://broadcastdialogue.com/canadian-news-publishers-sue-anthropic-over-ai-copyright-infringement/
• NDTV Profit, “Anthropic’s 1.5 Billion Book Settlement Leaves A Wider Copyright Fight Ahead,” 2026: https://www.ndtvprofit.com/markets/anthropics-1-5-billion-book-settlement-leaves-a-wider-copyright-fight-ahead-12240505/amp/1
• Adweek, via reporting on 4 November 2026, on the reported Anthropic and Associated Press licensing deal