On October 17, 2026, Meta removed a sentence from its updated commercial terms of service that legal experts read as a free license to scrape publishers' websites and train AI models on them. The clause had an effective date of October 30, 2026, no opt-out, and a choice between accepting it or abandoning Meta's platforms. Its deletion β first reported by The Verge and confirmed by TechCrunch β marks the licensing wave's second act.
The clause and the backlash
Meta's updated commercial terms, which govern businesses using Facebook, Instagram, WhatsApp, and Messenger commercially, authorized automated tools such as scrapers, bots, and crawlers to access content from websites "associated with" a business account. The explosive sentence was the one that followed:
"Meta may use any content from such domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of, the Meta Company Products, services, and features."
The word "train" β paired with a definition of the covered products that included "AI at Meta" β made the AI-training angle explicit. Nieman Lab had three legal experts review the terms. Matt Topic (Loevy & Loevy) called it "a remarkable provision to bury in terms of service, given that this would give Meta access to material that would likely cost many millions of dollars in a licensing market." James Grimmelmann (Cornell) said the language "clearly represents permission to Meta to train AI on any content from their sites and possibly to deploy some of that content, even in competition with the publishers themselves." The News/Media Alliance called the terms "outrageous."
The removal β and what Meta said about it
Meta first insisted the update reflected no change in how it crawls the web; communications VP Andy Stone said it was about accurate product listings for ad recommendations, "not about scraping news sites." Then, on October 17, the contested language disappeared. Meta's statement to The Verge: "We've heard you and want to clarify: the crawling terms we consolidated only applied to our ads and business products, and we've clarified that in the terms themselves."
Note what the removal changed and what it didn't. It changed the contract β the written authorization to "train" is gone. It did not change Meta's crawling behavior, which the company says was never altered. The central unanswered question β whether the terms could have overridden a site's robots.txt signals β stays unresolved: Meta says sites "can use robots.txt to signal their preferences" and that it gained no "new rights that supersede those preferences," but that assurance rests on a statement, not contract language.
Act one: negotiated licenses. Act two: the ToS grab.
The episode is the licensing wave's second act. Act one was the negotiation phase: AI companies cutting paid content deals with publishers β Meta itself holds paid training-content deals with CNN, Fox News, USA Today, and People Inc., and the ReachβGoogle licensing template (covered October 16) has become a reference point for how those deals get structured.
Act two is the unilateral phase: instead of paying, rewrite the platform's terms of service to capture the same data by contract β free content for everyone except the largest publishers. Nieman Lab drew the sharpest contrast: Meta had just launched Meta One, a subscription charging businesses and creators to include links in posts and reels β Meta could have been charging for the same links it crawled for free.
What it means for content startups and data moats
For founders building content businesses, three assumptions changed:
- Your ToS counterparty is also your data counterparty. Platforms read their terms as data contracts even when you read them as advertising contracts. Read both meanings everywhere.
- The robots.txt question is unresolved, not settled. Experts were split on enforceability β the most honest summary. The first litigated case will set precedent.
- Data moats have a contractual attack surface. Crawling hygiene β robots.txt and bot-management posture, documented licensing preferences, knowing which accounts link to which domains β is now core infrastructure.
The two-tier licensing market: what investors should read
Zoom out and a market structure emerges. The biggest publishers get paid licenses. Everyone else β the mid-tier sites, niche verticals, business blogs β gets their content captured through platform terms, aggressive crawlers, or default access. That is a two-tier licensing market: negotiated compensation at the top, unilateral extraction below.
For investors, three reads follow. First, content businesses with scarce, high-quality data are acquisition targets β training-data pricing is being discovered in real time, and Topic's "many millions of dollars" is a useful anchor. Second, startups whose training pipelines depend on open-web crawling carry terms-of-service risk that rarely appears in diligence decks; ask about it. Third, the enforcement gap itself is an investment thesis: tooling for crawler permissions, content licensing, and attribution is selling into a market this episode just advertised.
Practical takeaways
- Audit your platform accounts before October 30: which accounts link to which domains?
- Tighten crawling hygiene: explicit robots.txt policies for AI crawlers, bot-management rules, a documented licensing position.
- Price your data: publishers that know their training-data value will negotiate; those that don't get defaulted into free.
Scrutiny worked β a clause deleted within days. But the retreat was tactical, not philosophical. The second act is about who writes the terms, and that fight is not finished.
Angeled.com connects founders building in data, content, and AI infrastructure with investors.
Read Next
Open Weights Are Not Open Business Models: Mistral Large 4 and the Licensing Trap
Courts Are Writing AI's Rules Now: Three Rulings Founders Should Read
Delhi High Court Orders 24-Hour Takedown of AI Deepfake Videos Targeting Chief Minister
EU Court Adviser Says AI Training Is Legal β but Gemini's Answers May Still Infringe Copyright
AI Regulation After the Midterms: What Founders Should Do Before the January Bills Arrive
DOJ Launches Super Intelligence Task Force to Police AI Misuse



