Meta Pulls AI Training From Commercial Terms, but Web-Crawling Rules Stay Broad

Meta quickly softened a planned change to its rules for business accounts. Owners had objected after reading a draft that appeared to clear the way for broad web scraping. Meta removed an explicit mention of AI training. Still, publishers say the remaining language leaves serious uncertainty about how much online material the company can collect automatically.

The first version came out the previous week and was due to take effect on October 30. Nieman Lab highlighted language that covered web scrapers, bots and similar systems. It said those tools could reach domains, pages and addresses linked to business accounts, then use what they found to build and improve Meta services. AI training was one of the listed purposes. The likely effect worried publishers. Most digital publishers have business accounts on Facebook or Instagram, so the draft seemed to place their entire sites within Meta's reach.

Meta then quietly changed the draft again. The reference to training disappeared. The new wording frames the purpose as supporting advertising and business tools, conducting research, and maintaining product integrity and security. It still permits automated software to visit, examine, gather and refresh sites connected to a customer's products or services, plus material available through people's use of Meta services. It also allows test credentials. In a statement, spokesperson Andy Stone called the edit a clarification. He said the clause applies only to advertising and business products, was not about collecting news, and does not change Meta's use of web crawlers.

That reassurance has not quieted all critics. Nieman Lab called the revised clause a substantial expansion of Meta's scraping permissions. Several legal experts who spoke to the publication said the rules could let the company crawl news sites with few restrictions. Stone said the change does not override a site's robots.txt preferences. Yet robots.txt carries no binding legal authority, and AI companies often bypass it. The tension sits against Meta's history with publishers. The company stopped paying for news in 2022, removed its dedicated news tab in 2024, and has recently made a small number of deals to supply news content to its AI chatbot.