The New York Times and three other outlets sued Brave over content it allegedly scraped and supplied to five AI companies, moving the fight from the model developers to the supplier that sells publisher content to them.
The New York Times, the Daily News, the Chicago Tribune and the Orlando Sentinel sued Brave Software in federal court in Manhattan on Monday, alleging it scraped their copyrighted content and distributed it to AI companies that are its customers. The suits name Anthropic, Perplexity, Mistral, xAI and Cohere.
The Times compares Brave to a supplier and distributor of stolen content for AI developers. It alleges Brave built an index of more than 30 billion web pages that publishers cannot block, runs a browser feature that bypasses paywalls, and blocks publishers’ ads while selling its own. The Times seeks an injunction and damages. Brave did not immediately respond to a request for comment.
Reporting on the filings adds that the cases carry five counts, including direct, contributory and vicarious copyright infringement and DMCA circumvention, and that a Times test returned 1,291 of 1,295 words of one story.
The suit moves the fight from the AI labs that train on publisher content to the supplier that sells it to them.
I learned this running embedded finance programs across 12 countries: when several parties are paid from one product, everyone has to read the same record of what moved. Publishers that keep their own record of who takes what will negotiate from facts, not claims, and that record is worth building before the first demand letter goes out, not after.
An adviser to the EU’s top court said a chatbot can infringe press publishers’ rights in two ways: by getting past a paywall or bot block, or by answering in a way that makes the original article unnecessary. The opinion, in a case brought by a Hungarian portal against Google, rejects the argument that freely accessible articles are implicitly licensed for chatbot reuse.
It says training on lawfully accessible works is in principle covered by the EU text and data mining exception, subject to a machine-readable opt-out such as robots.txt. The opinion is non-binding.
The Operator’s Take: The adviser draws the line publishers wanted: reuse that substitutes for the article is treated differently from learning from it. The opt-out matters as much as the infringement test, because it decides whether training is covered at all.
Legal · Industry Positioning Confirm your EU-facing terms and robots.txt carry a machine-readable training reservation, and record the date it went live.
People Inc. said Google search referral traffic has fallen 70% over two years, including 22% in the third quarter, and core sessions fell 26%. Digital revenue still rose 2% to $289 million. Revenue not tied to site sessions is now 43% of digital revenue and grew 15%, while licensing revenue rose 49%, led by deals with aggregators and AI vendors.
The company expects licensing growth to slow in the fourth quarter against a tough comparison with last year’s Meta deal. It cut about 6% of staff the week before.
The Operator’s Take: A career that began valuing broadcasting and publishing companies for M&A teams teaches that a growth rate is only as durable as the contract behind it. A 49% jump measured against a prior-year deal is not a run rate.
Commercial · Economic Model Rebuild your licensing forecast deal by deal, with term and renewal dates, and show the board licensing revenue separately from session-based revenue.
ROSS Intelligence has asked the full Third Circuit to rehear its AI copyright case with Thomson Reuters over Westlaw headnotes, and has hired appellate litigators from Gibson Dunn to lead the petition. ROSS argues the panel asked the wrong question on fair use, and says treating training on copyrighted material as infringement, regardless of how similar the output is, would halt AI development.
The panel’s opinion was published on September 30. A blog tracking the case reads the choice of counsel as a sign that ROSS is aiming at the U.S. Supreme Court.
The Operator’s Take: Whether or not the full court takes the case, the petition keeps one question open: does training alone count as infringement? Any license priced on today’s answer should be able to move if that answer changes.
Legal · Deal Execution Add a clause to every pending AI license that reopens price if a controlling court changes the fair-use standard.
Publisher trade groups, including the News Media Alliance, Digital Content Next and the UK’s Professional Publishers Association, are pressing Meta to change planned terms of service. Several said Meta had not yet contacted them. Some publishers say they may leave, but audience heads at three said they are not ready.
Facebook traffic to the top 10 publisher sites is down 5% year over year, and Meta averaged 3.47% of global publisher traffic in July. The UK trade body wants platform access unbundled from permission to crawl content, and no retroactive authorization of past copying.
The Operator’s Take: At 3.47% of traffic, Meta is small enough to leave and still big enough that no one has. The unbundling demand is the right one to standardize, because it separates reach from rights.
Advocacy · Industry Positioning Ask your trade body to press for terms that separate platform access from crawl permission and exclude retroactive authorization of past copying.
IFRRO, the international body for reproduction rights organizations, released a study of the legal, policy and market impact of AI on text and image works. It combines copyright principles, emerging case law and a 2026 global survey of nearly 2,000 respondents. More than 90% say generative AI has hurt rightsholder earnings, 96% say the current copyright framework does not adequately protect them, and more than 50% believe AI-specific licenses would greatly improve their economics.
The study concludes that large-scale use without consent, transparency or payment is neither legally nor economically sustainable. A webinar is set for November 23.
The Operator’s Take: A survey is not a deal, but it shows what collective licensing bodies will bring to the table. Expect buyers to meet the same three conditions from the collective side: control, copyright principles and transparency.
Commercial · Deal Execution Compare your direct-deal terms with the study’s three conditions and note where a collective license would undercut them.