Meta removes clause that raised concerns over AI training

Summary

The Verge and Nieman Lab highlighted a legally problematic clause in Meta's terms of service. This clause reportedly granted the company broad rights to scrape content from any publisher with a Meta business account for AI training purposes. Meta subsequently and quietly announced the removal of this clause.

Why it matters

This incident demonstrates that updates to terms of service can effectively expand data rights in the context of AI training, often without public awareness. It also underscores the ongoing legal and ethical tension between platform companies and content creators, even as Meta faces high-profile copyright lawsuits and large-scale model training continues.

Highlights

• Legal experts described the clause as a "notable provision," noting that merely owning a Meta business account could have allowed Meta to use a publisher's website content for AI training.
• Meta initially defended the change by claiming it did not reflect a change in how Meta crawls the web, but the company ultimately removed the relevant language entirely from the text.
• According to The Verge, Meta made this change after press scrutiny, suggesting the company acted under public pressure rather than proactively correcting an erroneous statement.

Critical take

Meta's post-incident claim that the clause applied only to ads and business products is unconvincing. The original text explicitly referenced "domain names, web pages, or URLs" associated with business accounts, which form the foundation of most publishers' online presence. The company's decision to remove the clause entirely rather than clarify it suggests the original wording was at least dangerously ambiguous and likely intentional.