Meta has deleted a controversial provision from its commercial terms of use that legal experts warned could have allowed the company to scrape content from publisher websites and use it for Meta AI training. The provision, removed on Saturday, October 17, had appeared in a revision of Meta’s commercial terms that was scheduled to take effect on October 30.
The deleted language stated that 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. While the word “AI” did not appear in the clause, experts interviewed by Nieman Lab said it clearly represented permission for Meta AI training on any publisher content from sites linked to a Meta business account.
Meta AI training fears raised by legal experts
The alarm was sounded by several legal scholars, including James Grimmelmann, a professor of digital and information law at Cornell. They argued that the provision was broad enough to give Meta a free right to use scraped content for Meta AI training and could even supersede blocking instructions in a site’s robots.txt file, the standard mechanism publishers use to keep crawlers out.
Grimmelmann told Nieman Lab, as reported by The Verge, that the terms would let Meta train AI on content from publishers’ sites and possibly deploy some of that content in competition with the publishers themselves. The concern applied to any business with a Meta business account across Facebook, Instagram, WhatsApp or Messenger.
Meta AI training stance: company pushes back
Meta initially pushed back, with a spokesperson insisting the update did not change how the company crawls the web. Communications VP Andy Stone said on X that the change was intended to pull accurate product listings for advertising recommendations and was “not about scraping news sites.”
On October 17, however, Meta removed the clause entirely. A spokesperson said, “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.”
What the revised terms actually say
Under the revised terms, automated crawling remains but its stated purposes are narrower: the collection of website content is now linked to developing, improving and researching advertising and business products, plus maintaining the security of Meta’s products. The explicit reference to “train” has been removed.
Nieman Lab’s comparison of the two versions noted that the revised wording does not explicitly restrict the use of collected information in every AI-related application. Meta spokesperson Daniel Roberts, however, said website owners can continue using robots.txt to indicate their access preferences and that the changes do not alter existing crawling practices.
Meta AI training and the copyright wars
The episode is the latest skirmish in the broader battle between AI companies and publishers over training data. Meta defeated a high-profile copyright infringement lawsuit last year, but proceedings against AI companies, including Meta, continue. Publishers argue that AI systems are being built on copyrighted news content without permission or compensation, while the companies say training on publicly available material is fair use.
The swift U-turn shows how sensitive the issue has become. With Meta AI training ambitions spanning its Llama models and assistant products, even a commercial-terms update can trigger a major legal and public-relations firestorm within days.
Publishers stay alert after the reversal
The revised terms are scheduled to take effect on October 30, and publishers will be watching closely. Meta spokesperson Daniel Roberts said the changes do not alter existing crawling practices and that website owners can continue using robots.txt instructions to indicate their access preferences, while communications VP Andy Stone rejected suggestions the update was ever meant to enable scraping of news websites.
But the episode has sharpened publisher vigilance. Nieman Lab’s comparison of the two versions noted that the revised wording still does not explicitly restrict the use of collected information in every AI-related application, and legal experts say any future Meta AI training on commercial-terms data would face intense scrutiny after this U-turn. The swift reversal shows that publishers and regulators now treat every clause as a potential Meta AI training gateway.