Meta updated its not-yet-effective new commercial terms again on October 17, 2026, and the key sentence that critics said opened the door to AI scraping has disappeared from the terms page. The Verge reported the same day that the sentence stating Meta may use any content obtained from associated domains, pages and URLs to provide, develop, improve, research, train and maintain its products no longer appears on the commercial terms page. The terms are due to take effect on October 30; only days into the backlash, Meta has removed the most conspicuous sentence first.
The deleted sentence is precisely the one mentioning training
That sentence became the focus because the uses it listed included training. As covered in the earlier dispute over Meta's new commercial terms, legal experts and observers such as Nieman Lab had pointed out that the wording covered news websites associated with business accounts, and that Meta might later rely on it to claim a right to scrape that content for AI training, bypassing content licensing negotiations; James Grimmelmann, a professor of digital and information law at Cornell, among others, worried it could be read as overriding the refusal signals websites express through robots.txt. Now the sentence that was singled out has vanished from the text before the effective date even arrived.
Meta calls it a clarification, not a change of course
Meta spokesperson Daniel Roberts explained the edit by saying the company had heard the feedback and wanted to clarify its intent: the consolidated crawling terms apply only to advertising and business products, are not about scraping news sites, and give Meta no new rights that supersede site owners' preferences; content owners can still use robots.txt to signal their choices as always, and the way Meta uses its crawlers has not itself changed.
This line is consistent with Meta's response in the previous round: from the start, the company said the update merely consolidated wording scattered across several documents into one place, without changing actual practice. The difference is that last time the clarification was verbal and the text stayed put; this time the text actually changed. For a company accustomed to publishing terms first and gauging reactions later, publicly deleting a disputed sentence before it takes effect shows at minimum that concentrated criticism from publishers and the legal community produced real pressure.
What did not change after one sentence was deleted
It is worth being clear about the scope: what was removed is one authorising sentence, not the terms as a whole. The October 30 effective date has not been postponed, the framework authorising crawling for advertising and business products remains, and Meta states explicitly that how its crawlers actually operate is unchanged. In other words, the scraping reality websites face today has not changed because of this edit; what changed is that Meta now holds one less piece of text it could cite — if a scraping dispute arises later, the textual basis for arguing that business users already consented to their website content being used for training has been withdrawn by Meta itself in advance.
For publishers and businesses holding commercial accounts, the actionable conclusions still sit on their own side: robots.txt settings remain the main tool for expressing crawling preferences, and Meta has again confirmed their effect; and it remains worth auditing, before October 30, which domains are associated with one's business accounts and could fall under the terms. What to watch next is whether similarly broad wording survives in Meta's other product terms, and whether actual crawling behaviour after the effective date matches this clarification — deleting text settles the paper authorisation, but the real test lies in enforcement after the terms take effect.
