Meta Removes Language Allowing News Sites Scraping for AI Training
Meta has removed a controversial clause from its commercial terms that had raised concerns among publishers and legal experts about the company's ability to scrape content from third-party websites to train artificial intelligence models. The passage, which was located on the company's "Commercial Terms" page rather than the user-facing privacy policy, authorized Meta to "use any content from domains associated with business accounts" to "provide, develop, improve, research, train, and maintain the integrity" of Meta's products.
The passage was removed after a rapid response by The Verge, which identified the language and published a report on October 17. TechCrunch, on the same date, published a detailed report on the legal impact of the clause and reactions from experts. The Nieman Law Report, in a report dated October 3, had warned that the clause, which was scheduled to take effect on October 30, could give Meta permission to scrape content from any publisher with a business account on the platform.
What was in the text
The original passage was not in the end-user privacy policy section, where personal data collection terms are normally found. It was in the "Commercial Terms" page, aimed at businesses using Meta's platforms for marketing, sales, and customer service. The company's logic was clear: if a business posts content on Meta's social networks, Meta itself should have broad rights over that content to serve ads and commercial products. The problem is that the text did not limit itself to content published directly on the platform. The passage stated: "Meta may use any content from such domains, web pages or URLs to provide, develop, improve, research, train, and maintain the integrity of the company's products". The word "such" referred to domains associated with business accounts, which meant in practice that any website linked to a commercial page on Facebook, Instagram or WhatsApp could have its content collected to train AI models.
James Grimmelmann, a professor at Cornell Law School, told Nieman Lab that the passage gave Meta the right to scrape content from news sites without payment, bypassing the robots.txt and content licenses that publishers had already negotiated with other platforms. "This is not how copyright works," Grimmelmann said. "If you want to use someone's content, you need to negotiate a license or prove you have a legal right. This is not a legal right that exists today."
Industry response
Nieman Lab published, on October 3, a prominent alert to editors about the change. The article pointed out that the passage went beyond what had been Meta's established practice. "Traditionally, you have this idea that Meta might use content [posted] on Facebook or Instagram for training," Beale, a company representative, told Nieman Lab. "But this was different because they're saying that since you are a business user of Facebook, they have the right to go out and grab information from basically wherever you have a presence."
The response was immediate. News organizations such as the News Media Alliance, which represents hundreds of publications in the UK and Europe, published notices alerting their members. PCMag reported on October 16 that the change had already generated concerns in at least three countries. OECD.AI logged the incident as an intellectual property rights violation case, in the "harm" category related to AI.
Meta itself did not give a technical explanation for the inclusion of the passage. Andy Stone, a company spokesperson, told The Verge: "We've heard you and want to clarify: the scraping terms we consolidated only applied to our ads and business products, and we've clarified that in the terms themselves". The statement, however, did not answer the fundamental question: what is the legal basis that Meta now uses to train its AI models on third-party content?
What changes after the revision
The revision of the terms, made on October 17, removed the passage that explicitly mentioned "training" of AI from business domains. In the new text, the company states that the scraping terms apply only to "ads and business products" and that "they do not change the way Meta uses web crawlers". The revision is a tactical victory for publishers, but it does not resolve the core issue of data collection. Meta has not yet clarified what is the legal basis it uses to collect content from external websites to train its AI models. The company has been sued by several publishers for unlicensed content scraping, and a legal action against Meta has been ongoing since the second quarter of 2026, when a group of European publishers filed a class action against the practice.
The question that remains is: if Meta can no longer use commercial terms to scrape business domains for AI training, what legal basis does it invoke to continue collecting content from external sites? The answer could determine not only the future of the relationship between social media platforms and publishers, but the very business model of hundreds of publications around the world.
Sources: The Verge, TechCrunch, Nieman Lab
✓ Independent sources cross-checked and verified before publishing
