Meta added language to its business account terms saying it could use automated scrapers to collect content connected to a business's Facebook, Instagram, Messenger, or WhatsApp account and use that content to train its AI products, under terms set to take effect October 30, 2026. Meta removed the clause after legal experts and the publication Nieman Lab raised concerns about it, TechCrunch reported October 17.
Meta's Updated Terms Covered Any Domain Linked to a Business Account
The update applied to business accounts managing Facebook Pages, Messenger, Instagram profiles, or WhatsApp channels, including accounts run by news publishers. Its language authorized Meta, or any entity acting on its behalf, to use automated software such as web scrapers, bots, and crawlers, and to create test credentials, to access, index, cache, analyze, collect, retrieve, or crawl the domains, web pages, or URLs associated with that account, and to use any content found there to "provide, develop, improve, research, train," and maintain Meta's products. Businesses could not opt out before the October 30 effective date; the terms said accounts that disagreed had to stop using Meta's products for business purposes entirely.
Lawyers Call the Clause a 'Remarkable Provision to Bury'
Matt Topic, a media attorney at the law firm Loevy & Loevy, told Nieman Lab the language was a "remarkable provision to bury in the terms of service." He said it could hand Meta content that would otherwise cost millions of dollars to license, and that while it is unclear whether a court would enforce the clause, publishers should assume that it would be.
Cornell's James Grimmelmann Says the Clause Could Build an AI Archive of News Sites
James Grimmelmann, a law professor at Cornell, told Nieman Lab the broadest reading of the clause could let Meta train AI on any content from a news site's connected domain and potentially use that content to compete with the publisher directly. He also flagged the clause's "test credentials" language, saying it could let Meta "sign up as a fake user" to get past paywalls or registration walls. Grimmelmann summed up publishers' position as having to trust that Meta would not use the full extent of what the terms allow.
Meta Says the Update Only Consolidates Existing Ad Language
A Meta public affairs spokesperson told Nieman Lab that crawling language had been scattered across several of the company's terms, including terms covering generative AI in ads, and that the update simply consolidates that language in one place. The spokesperson said the new terms do not change how Meta crawls publisher sites, and that site owners can still manage access through robots.txt. Meta declined to answer whether its "test credentials" language could be used to bypass paywalls, and declined to comment on Meta One, a separate program that limits external links for some accounts.
TechCrunch Reports Meta Pulled the Clause Within Days
Nieman Lab published its findings October 15, and TechCrunch reported October 17 that The Verge had since reported the scraping provision was removed from Meta's terms. The reversal came as Meta continues to face scrutiny over how it trains its AI models: the company won a high-profile copyright lawsuit over AI training on books last year, while other lawsuits against Meta and other AI companies over training-data practices remain active in US courts.
Why This Matters for Engineering Leadership
Terms-of-service language is becoming a direct lever in AI data rights, and this episode shows how quickly that language can change under public pressure, in this case within about 48 hours of Nieman Lab's report. Engineering and legal teams that integrate with Meta's business tools, or with any platform whose terms grant broad rights to scrape and train on connected content, should treat terms-of-service updates as a recurring review item rather than a one-time signature, and should keep dated records of the exact clause language in effect at any given time. The same caution applies in reverse: any engineering organization building its own AI products should expect its customers' legal teams to start reading data-use clauses in commercial terms far more closely than they have in the past, and should write those clauses to survive that scrutiny rather than counting on them going unread.
FAQ
What did Meta's updated business terms say about AI training?
The terms said Meta could use automated scrapers to collect content from domains connected to a business account and use that content to "provide, develop, improve, research, train," and maintain Meta's products, including its AI systems.
Did Meta remove the AI training clause?
Yes. TechCrunch reported October 17, 2026 that The Verge found the scraping provision had been removed from Meta's business terms after Nieman Lab's report drew attention to it.
Why did lawyers call Meta's clause a "remarkable provision to bury"?
Attorney Matt Topic said the clause could give Meta access to content that would otherwise cost millions of dollars to license, buried inside routine terms of service that most account holders would not read closely.
Could Meta's clause have let it bypass paywalls?
Cornell law professor James Grimmelmann said the clause's language authorizing "test credentials" could potentially let Meta sign up as a fake user to get past registration or paywall barriers, though Meta declined to answer questions about that use directly.
What did Meta say about the updated terms?
A Meta spokesperson said the update only consolidated crawling language that had been scattered across several existing terms, including AI-in-ads terms, and that it does not change how Meta crawls publisher sites.
Has Meta faced other AI training lawsuits?
Yes. Meta won a high-profile copyright lawsuit last year over training AI on books, and other lawsuits over its AI training practices remain active in US courts.