Meta has removed language from its updated commercial terms that legal experts said could have given the company permission to scrape the websites of businesses, including news publishers, and use their content to train its AI models. TechCrunch reported the removal on October 17, following reporting by The Verge that the provision no longer appears in the terms.
The reversal comes days after GEroundtable reported the original clause in Meta’s new business terms, which are scheduled to take effect on October 30. Meta says the change clarifies the terms and does not change how its web crawlers actually work.
The Clause Meta Has Removed
The disputed provision 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 terms also authorised automated tools, including scrapers, bots and crawlers, to access and collect content from websites associated with businesses using Meta’s platforms.
Legal experts interviewed by Nieman Lab called the wording a remarkable provision. Cornell digital law professor James Grimmelmann said the terms clearly represented permission for Meta to train AI on any content from those sites, and possibly to deploy that content in competition with the publishers themselves. Critics also argued the clause could have overridden the preferences site owners express in robots.txt, the standard file sites use to signal which crawlers may access their content.
That reading mattered because of who the terms cover: any publisher or business with a Meta business account across Facebook, Instagram, WhatsApp or Messenger could have been swept in, without a separate licensing negotiation of the kind described in our explainer on Meta AI and Llama.
Meta’s Response: A Clarification, Not A Change In Crawling
Meta disputed the interpretation from the start. A company spokesperson insisted the updated terms did not reflect a change in how Meta crawls the web, and communications vice president Andy Stone said the update was meant to pull accurate product information for advertising and content recommendations, not to scrape news sites.
After removing the provision, Meta 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. It does not change how Meta uses web crawlers.”
Meta also told The Verge that content owners can use robots.txt to signal their preferences, and that the update does not give the company new rights that supersede those preferences. The ANews report on the revision adds that spokesperson Daniel Roberts said the changes do not alter Meta’s existing crawling practices. The distinction between contractual permission and stated practice is the heart of the dispute, and it is why Meta’s open-weight strategy, covered in why Meta gives its AI models away, does not settle how it obtains training data.
What Remains In The Terms From October 30
The removal changes the contract, not the crawling. Provisions authorising automated access to business-linked websites for advertising and business products remain in place, and the revised terms still take effect on October 30. What is gone is the sentence that extended the stated use of collected content to training Meta’s products.
That leaves publishers in the position they were in before the update: relying on robots.txt signals, public statements and, for the largest outlets, paid licensing deals. Meta has licensing agreements with publishers including CNN, Fox News, USA Today and People Inc. for AI-related content, deals the deleted clause would have made unnecessary for any other business-linked site.
Meta defeated a high-profile US copyright lawsuit over AI training data last year, while other cases against AI companies, Meta included, continue. The episode therefore closes one contractual route, not the wider legal fight, a pattern familiar from the copyright fights reshaping AI.
Why It Matters
The central entity is Meta’s commercial terms, and the reversal matters for every business that both depends on Meta for distribution and publishes content Meta’s AI could learn from. For a few days, a single sentence in a terms update appeared to convert a business account into a training-data licence, free of charge and without a separate opt-in. Its removal, within days of expert criticism, shows that terms-of-service language is now a front line of the AI data fight, read as closely by publishers and lawyers as any model release. Businesses should take the practical lesson: the document worth re-checking after any platform dispute is the terms page itself, not the press statement about it.
Background
Meta, the parent of Facebook, Instagram and WhatsApp, trains its Llama models and its Meta AI assistant on large volumes of web and platform content, and has defended that practice in court while also signing paid deals with major publishers for AI features. Its commercial terms govern the businesses that advertise and operate across its apps, and an updated version consolidating crawling provisions was published ahead of an October 30 effective date. When Nieman Lab and legal experts flagged the training language in mid-October, the story spread quickly through the publishing industry, which is already in litigation and negotiation with AI companies over the value of its archives, and Meta removed the provision on October 17.
Source: TechCrunch, October 17, 2026; ANews, October 18, 2026
0 Comments