Meta removes controversial AI scraping language from commercial terms

Meta has removed specific language from its commercial terms that legal experts argued permitted the company to scrape content from news sites for artificial intelligence training. The update follows significant backlash from media organisations and legal scholars who believed the previous wording granted Meta broad rights to use external content without permission. The revised terms no longer contain the clause that suggested Meta could utilise content from domains and web pages to develop, improve, and train its products and services.

The removed text had been interpreted by figures such as James Grimmelmann, a professor of digital and information law at Cornell, as an attempt to supersede blocking instructions found in site robots.txt files. This interpretation suggested Meta intended to use scraped data for free to train its AI products. In response to the criticism, Meta initially maintained that its use of web crawlers had not changed and that any contrary suggestions were false. However, the company has now clarified the scope of its crawling terms to limit them to ads and business products.

In a statement to The Verge, spokesperson Daniel Roberts confirmed that the updates do not grant new rights that override content owner preferences. He stated that content owners can continue to use robots.txt to signal their preferences regarding data collection. The company emphasised that the changes were intended to clarify the application of the terms rather than to alter its fundamental approach to web crawling or data acquisition for its AI initiatives.