Meta backs down from scraping professional sites for AI training

Meta planned to include in its updated terms of service, effective from October 30, a clause allowing automated scraping of content from websites linked to its business users. This content could have been used to train the group’s AI models, raising concerns among publishers and other organizations managing Facebook or Instagram pages.

The updated terms would have permitted Meta to use bots to explore, index, and collect web pages associated with professional accounts to develop or improve its products. The company argued that owners could still control access via the robots.txt file. However, legal experts warned that these terms granted access to content normally protected by expensive licenses.

Following strong backlash, Meta clarified that automated collection was limited to its advertising and commercial products, not AI model training. The company also removed the term “train” from the terms, refocusing usage on advertising and user engagement. GESTE, a publishers’ group, criticizes this widespread practice among major platforms, which impose unfavorable contracts on media outlets.