Meta has removed controversial language from its updated business terms that appeared to grant the company permission to scrape content from publishers' websites and use it to train artificial intelligence models, following sharp criticism from legal experts and media industry observers.
The clause, included in Meta's newly consolidated commercial terms of use, declared that the company 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's products and services. For more context on this story, see our ongoing artificial intelligence updates.
A 'remarkable provision'
Legal experts interviewed by Nieman Lab, the journalism-focused publication of Harvard's Nieman Foundation, described the wording as a "remarkable provision" that could have been read as permission for Meta to crawl the website of any publisher with a Meta business account and feed that content into its AI training pipeline.
The language arrived at a sensitive moment. Meta defeated a high-profile lawsuit over alleged copyright infringement in its AI training last year, but other cases against AI companies, including Meta, continue to move through the courts. Publishers have spent the past several years fighting scraping-based AI training in courtrooms and through licensing negotiations, which is precisely why the new wording drew fire within hours of being flagged.
Commentators also noted a mismatch between Meta's existing crawler behavior and the breadth of the clause. Meta's crawlers have historically fetched pages to render link previews and power business tools, not to assemble training corpora. The new terms, critics argued, would have quietly redrawn that line.
Meta: language 'clarified,' behavior unchanged
The provision has now been scrubbed from the terms. As The Verge reported, and TechCrunch confirmed, Meta issued a statement acknowledging the reversal.
"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 company said. "It does not change how Meta uses web crawlers."
A Meta spokesperson had previously insisted that the updated terms did not reflect any change in the way Meta crawls the web, framing the language as a consolidation of existing policies rather than an expansion. The removal of the clause suggests the company recognized that the wording said something considerably broader than it intended.
The episode is a case study in how much weight now rides on individual sentences in platform agreements. Phrases like "train" and "research" have taken on new commercial significance in the AI era, and terms that were drafted to cover advertising tools can read very differently once AI model development enters the list of permitted uses. Meta's own explanation — that crawling terms had merely been consolidated from ads and business products into the general agreement — is precisely the kind of drafting detail that most businesses never see until someone flags it.
According to multiple reports, the updated commercial terms are scheduled to take effect on October 30, which gave businesses a fast-approaching deadline to accept the agreement as written. The revision lands before that cutoff.
Experts remain cautious
While Meta's walk-back addresses the most alarming reading of the clause, legal observers urged publishers not to consider the matter closed. Follow-up coverage from Nieman Lab noted that legal experts believe the revised terms may still grant Meta broad permissions to scrape website content, even with the AI-training language gone.
That concern reflects a structural problem in platform economics: businesses that depend on Facebook, Instagram, or WhatsApp for marketing and customer communication have little practical choice but to accept Meta's terms, whatever they say. The leverage in the relationship runs one way, and fine print rarely gets negotiated with small publishers.
Why publishers are on edge
The episode fits into a larger fight over who profits from online content in the AI era. News organizations have watched search referrals decline as AI assistants answer questions directly, and several have signed licensing deals with AI developers while others litigate. Against that backdrop, the prospect of a platform company granting itself automatic training rights over its business customers' content struck many as a step too far.
It also comes as regulators scrutinize how AI companies acquire data. In the European Union, enforcement of the AI Act's transparency requirements is ramping up, and training-data provenance is a central theme. In the United States, courts are still sorting out how copyright law applies to models trained on copyrighted works. Whether a terms-of-service agreement can even grant rights over third-party content — material the business customer does not own, such as wire-service articles or guest posts by freelance writers — is another unresolved question legal commentators have raised in similar disputes.
For now, the industry is watching whether other platforms attempt anything similar. Automated crawling for AI training has become a flashpoint across the web, with publishers increasingly blocking known AI crawlers and AI developers negotiating directly for licensed access. A clause that would have flipped that dynamic by default — enrolling every business customer's website into training access through a standard agreement — would have been a significant escalation, which helps explain why the backlash was so immediate.
What businesses should watch
For publishers and businesses with a Meta presence, the practical takeaway is to read the October 30 terms before accepting them. Consolidated agreements routinely reorganize old language, and as this episode shows, subtle rewording can carry major implications for how content may be used.
Meta's quick reversal also demonstrates that public pressure can still move platform policy, at least when the proposed change is unambiguous enough to spark a unified reaction. Whether the company's crawlers behave any differently than before is, as both Meta and its critics acknowledge, harder to verify from the outside — which is exactly why the experts are still reading the fine print.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
