Meta has deleted a controversial passage from its commercial terms of use after legal experts warned it could have allowed the company to scrape news websites and use their content to train AI models. The clause vanished from the terms page on Saturday, October 17, just two days before the new terms were set to take effect.
The removal came after a week of escalating criticism from publishers, media trade groups and copyright lawyers, who argued the wording was broad enough to hand Meta free training data for its AI products while overriding publishers’ own blocking instructions. Meta insists its crawling practices have not changed.
What the clause actually said
The disputed passage appeared in an update to Meta’s commercial terms that was scheduled to kick in on October 30. It said Meta could use any content from a business account’s “domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of” its products and services.
Notice the missing word: the clause never said “AI.” But it did say “train,” and in an industry currently fighting over every word that can justify model training, that was enough to set off alarms. The terms applied to anyone using Meta’s business or commercial tools, which covers news publishers, brands, public figures and small businesses managing Facebook pages, Instagram profiles, Messenger accounts or WhatsApp channels.
Why publishers were furious
Nieman Lab flagged the provision first, and the legal commentary that followed was blunt. James Grimmelmann, a professor of digital and information law at Cornell, called the language “absurdly broad.” Jason Kint, who heads a digital publisher trade group, wrote that the terms reached “outside its walled gardens.” Theo Bamber, CEO of a UK news media trade group, put a figure on what was at stake, calling the data involved worth “millions, or billions” in a licensing market.
The experts’ central fear was simple: the wording looked like permission for Meta to scrape the website of any publisher holding a Meta business account and feed the content into its AI models, possibly even in ways that compete with the publishers themselves. Worse, it appeared to supersede robots.txt files, the long-standing web standard publishers use to tell crawlers to stay out.
- Free training data: Meta could have collected publisher content without paying for a licence.
- Blocking bypass: The wording seemed to override robots.txt instructions that block AI crawlers.
- No real opt-out: The only way out was to stop using Meta’s products for business entirely, which is not a choice for publishers whose audiences live on Facebook, Instagram and WhatsApp.
- Fake sign-ups: The terms also let Meta create “test credentials,” which legal experts said could technically allow it to sign up as a fake user to get past registration walls or soft paywalls. Meta declined to say whether the credentials could bypass paywalls.
Meta’s defence, then the U-turn
Meta’s first response was to push back. Communications vice president Andy Stone said on X that the update was about pulling accurate product listings for ad recommendations, and was “not about scraping news sites.” A company spokesperson added that the update did not reflect any change in how Meta crawls the web.
That defence did not settle the matter. On Saturday, Meta deleted the contested language outright. Spokesperson Daniel Roberts told The Verge: “We have heard you and want to clarify: the crawling terms we consolidated only applied to our ads and business products, and we have clarified that in the terms themselves. It does not change how Meta uses web crawlers.”
In other words, the clause is gone, but the company says nothing substantive ever changed. Publishers, for their part, are treating the deletion as a win for close reading.
The bigger battle over AI training data
This episode is one front in a much larger war. Across the industry, AI companies are scraping publicly available, copyright-protected material from every corner of the internet to train their models. Elon Musk’s xAI uses public posts on X to train its Grok chatbot, while Meta scrapes publicly shared content from users above 18 for AI training, both without asking for consent first.
The legal ground is still shifting. Meta defeated a high-profile copyright lawsuit last year, but similar cases against Meta and other AI companies are still running. Meanwhile publishers are pushing for licensing deals, arguing that news content has real commercial value in the AI training market. Meta’s deleted clause shows why they are nervous: terms of service can quietly redraw the boundaries of what counts as permission.
What this means for Indian publishers and businesses
The story matters in India too. Indian publishers, digital newsrooms and businesses all use Meta’s commercial tools to reach audiences on Facebook, Instagram and WhatsApp, and many run advertising through the same accounts the deleted clause covered. For them, the episode is a reminder to read the fine print of platform terms as carefully as they read their own analytics, because a clause written in California can quietly affect a newsroom in Delhi.
For readers, the takeaway is simpler: the AI tools getting smarter every month are being trained on news you read for free, and the fight over who gets paid for that training is only heating up. Expect more such clashes as AI companies hunt for ever more data and publishers tighten their gates.
FAQs
What exactly did Meta remove from its terms?
A clause that said Meta could use any content from a business account’s domains, web pages or URLs to “provide, develop, improve, research, train, and maintain the integrity of” its products. Legal experts said the word “train” suggested it could cover AI model training.
When did Meta remove the clause?
On Saturday, October 17, 2026, after Nieman Lab and legal experts flagged it earlier in the week. The updated terms were due to take effect on October 30.
Did Meta admit it was going to scrape news sites?
No. Meta said the crawling terms only ever covered its ads and business products and that its web crawler practices are unchanged. It deleted the language to clarify, not as an admission.
Why did publishers care so much?
Because the clause seemed to override robots.txt blocking instructions, offered no opt-out short of leaving Meta’s platforms, and could have handed Meta free AI training data from publisher sites worth millions in a licensing market.
Is Meta the only company doing this?
No. Across the industry, AI companies scrape publicly available web content to train models. xAI trains Grok on public X posts, and Meta scrapes publicly shared content from users above 18, all without obtaining consent first.
That is all from the tech desk. Stay with Informeia for the latest in technology news.
