Meta Removes 'Train' from New Commercial Terms, But Legal Experts Warn AI Development Rights Remain Ambiguous

U.S. tech company Meta urgently revised its new commercial service terms on October 16, following strong backlash from the news publishing industry. The original terms, set to take effect on October 30, explicitly allowed Meta to use automated tools to scrape websites associated with business accounts and use the content to 'train' its products. This was widely interpreted as granting Meta free access to vast news databases for AI development.

According to reports by Nieman Lab and The Verge, the revised terms removed the word 'train' and restricted the use of acquired content to 'advertising and business products'. A Meta spokesperson stated that the move was intended to consolidate scattered crawler rules and did not change actual usage or override website robots.txt protocols. However, legal experts remain skeptical. James Grimmelmann, a professor at Cornell Law School, noted that Meta has never narrowed the scope of 'associated with your products or services', meaning its crawlers could still potentially access entire websites.

A more critical dispute centers on the definition of 'develop'. Media and intellectual property lawyer Matt Topic analyzed that the new terms still allow Meta to use scraped content to 'develop' business products, while 'business products' remains undefined. If Meta launches AI commercial services, the term 'develop' could legally be interpreted to cover model training processes. Additionally, Ars Technica discovered that the new terms include permissions to create 'test credentials', which could enable Meta's crawlers to bypass paywalls and access deep content.