Meta buried a permission slip for AI training inside its business paperwork. Then publishers and lawyers found it, and on Saturday, Meta quietly deleted it.

The company removed a provision from its updated commercial terms of use, set to take effect October 30, that had 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” Meta products and services. The Verge first reported the edit, and TechCrunch confirmed it.

All the heavy lifting in that sentence is done by the word “train.” It never appeared alongside “AI,” but three legal experts who reviewed the terms for Nieman Lab read it as a clear authorization to feed scraped website content into Meta’s AI systems, including Llama.

How one clause became a licensing shortcut

Here’s the mechanism that alarmed publishers. Nieman Lab reported on October 15 that under the updated terms, any business promoting content across Facebook, Instagram, WhatsApp or Messenger would automatically consent to Meta’s crawlers collecting content from websites “associated with” those accounts, simply by continuing to use the platforms.

The original language went further than ordinary crawling. It didn’t just permit collection; it listed training among the purposes. All three attorneys who reviewed the wording for Nieman Lab said it could let Meta argue it never needs a license to train on a publisher’s own site.

Matt Topic, a media and intellectual property attorney at Loevy & Loevy, didn’t mince words. “This is 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,” he told Nieman Lab. He added that while it’s unclear whether a court would actually enforce the clause, publishers should assume it would be.

Cornell professor of digital and information law James Grimmelmann put it even more directly. The terms, as quoted by The Verge, “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, which represents US news publishers, publicly called on Meta not to implement the change on October 30. That pressure mattered. It’s one thing for lawyers to grumble; it’s another when the industry’s trade group says so on the record.

Meta’s defense didn’t hold

Meta pushed back from the start. Communications vice president Andy Stone said on X that the update was meant to pull accurate product listings for ad recommendations, “not about scraping news sites.” A company spokesperson later insisted the updated terms did not reflect a change in how Meta crawls the web, and that the company respects robots.txt and other standard web protocols.

The problem with that defense was sitting right there in the text. If the clause was only ever about advertising and business products, why did it say “train”? And why did it cover “any content from” associated domains, rather than just product listings?

On October 17, Meta deleted the provision outright. A spokesperson, Daniel Roberts, told 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.” The statement added: “It does not change how Meta uses web crawlers.”

Independent comparisons of archived captures back that up. Unite.AI matched a terms preview from October 16 against one from October 17 and found the sentence listing training among the uses gone — the verb “train” no longer appears anywhere in the provision, and Meta’s live terms page matches the revised text.

What the deletion doesn’t fix

The crawling authorization itself survived the edit. The revised clause still permits automated access to domains, web pages and URLs associated with a business customer’s products or services. What’s changed is the stated purpose: now limited to advertising and business products, plus the integrity and security of Meta Company Products.

Read that carefully and you’ll notice what’s missing: a genuine opt-out. There is still no in-product mechanism that lets a business stay on Facebook and Instagram while removing its site from the scope of these crawling terms. The only complete way out is to leave the platforms — not a realistic option for most publishers.

The bigger picture

This is the third front in a fight that’s been escalating all year. AI companies need enormous amounts of training data; publishers need to be paid for it or at least to say no. Last year Meta defeated a high-profile copyright lawsuit over AI training data, but other cases against AI companies, including Meta, are still working their way through the courts.

What made this episode unusual is that it wasn’t a lawsuit or a bill. It was a few paragraphs of terms-of-service language, doing quietly what months of litigation try to do loudly: establish, by default, that a platform has the right to other people’s content. The clause was never even in force — the October 30 effective date hadn’t arrived — and it still forced a public retreat.

That’s the lesson publishers will take into every terms update from here on. The training data war isn’t only fought in courtrooms anymore. It’s fought in the fine print.

Sources: The Verge, TechCrunch, Nieman Lab