Meta removes controversial language allowing AI training on publisher content
Meta has removed terms of service language that appeared to grant the company permission to scrape and use business account content for AI model training.
What happened
Meta updated its commercial terms of use to state it could "use any content from [a business account's] domains, web pages, or URLs" to train and maintain its products. Legal experts characterized this as a "remarkable provision" potentially allowing Meta to scrape publisher websites and use content for AI training. After the provision drew scrutiny, Meta removed it, stating the language "only applied to our ads and business products" and claiming it did not represent a change in how the company crawls the web.
Context
The removal addresses concerns about Meta's content access for AI purposes at a time when the company faces ongoing copyright infringement lawsuits from creators and publishers related to AI model training. Though Meta previously defeated a high-profile lawsuit on this issue, other cases continue, making the perception of Meta's data practices significant to the company's legal and reputational standing. The clarity sought in the revision appears aimed at containing concerns that Meta's terms granted broad rights to use third-party content.
What's disputed
Meta's characterization of the removed language as not reflecting a change in practice; legal experts' interpretation of the original provision as a "remarkable" grant of scraping rights versus Meta's claim that it only applied narrowly to ads and business products.