Meta Rewrites AI Training Clause After Business Community Backlash

Quick Facts

• Meta removed language allowing AI training from its US commercial terms after one day of public criticism, but the original wording remained in UK and Korean editions as of October 19.
• The revised terms still permit Meta to crawl websites linked to business accounts — only the permitted uses of that content changed.
• Legal experts say the revised text may still allow AI training under the broader term "develop," which they argue is wide enough to cover model training.

Meta revised its pending commercial terms on October 17 to remove a sentence stating the company could use content crawled from business-linked websites to train its AI models, according to a review of the updated legal text. The change came within 24 hours of public criticism from legal scholars and news publishers.

The original terms, set to take effect October 30, would have automatically enrolled any business using Facebook, Instagram, or WhatsApp ads, pages, or groups into a consent framework allowing Meta crawlers to collect content from associated websites. The original language said Meta could use that content to "provide, develop, improve, research, train, and maintain the integrity of" its products.

Meta's previous commercial terms, in effect since March 2024, contained no crawling language. The entire crawling section is new, even after the edit.

What Changed

In its revision, Meta removed "train" from the list of permitted uses and narrowed the remaining uses to "advertising and business products" rather than all Meta products. The company said in a statement: "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 permission to crawl domains associated with business accounts remains in the text. Meta spokesperson Andy Stone wrote publicly that there was no change to how the company uses web crawlers and that content owners can use robots.txt flags to indicate their preferences.

As of October 19, the UK English edition of the terms still carried the original "train" language, as did the Korean edition. Meta had announced both changes — the deletion and the narrowing of scope — but neither had appeared in those editions.

Legal Experts Remain Skeptical

Media and intellectual property attorney Matt Topic said the revision falls short of a clean fix. "The revised version still allows Meta to use that scraped content to develop any 'business product,'" Topic said. "That doesn't seem to be a defined term, and presumably Meta is developing AI business products. I think 'develop' is broad enough to include 'train.'"

Cornell Law professor James Grimmelmann warned before the revision that the original terms represented explicit permission for Meta to train AI on publisher content and potentially deploy it in competition with publishers. He described the broadest interpretation as giving Meta a comprehensive archive of content it could serve directly or through AI.

University of Wisconsin professor Benjamin Sobel raised antitrust concerns, saying the structure of the terms echoes the legal concept of "tying." "If you want to use the Meta products that require acceptance of the commercial terms, you have to also acquiesce to the scraping and crawling permissions exacted in those terms," he said.

Publisher Response

The News/Media Alliance condemned the original terms before the revision. President and CEO Danielle Coffey called them "outrageous and probably illegal" and described the move as "an appalling, bad-faith policy" and "a clear abuse of the company's market power." The Alliance said the terms could not be allowed to take effect October 30.

Coffey's statement extended beyond news organizations. "News publishers aren't the only ones at risk — every publicly-hosted word, song, and image is up for grabs if Meta gets away with this," she wrote.

What It Means for Business Accounts

Any company running ads, managing a Facebook page, operating a WhatsApp business account, or using Meta Business Suite falls under the commercial terms. The terms offered no opt-out mechanism — continued platform use constitutes consent.

Topic noted before the revision that the provision could give Meta access to material worth millions of dollars in a licensing market. He said it was unclear whether a court would enforce the terms but advised publishers to assume enforceability.

The revised terms still take effect October 30. Meta has not announced changes to that date or to the crawling permission itself.