Meta's new terms say if its AI makes a mistake, it will be the customer's fault

Meta is rolling out revised commercial terms effective October 30 that shift legal risk for AI-generated mistakes onto enterprise customers and extend liability to actions by partners and contractors. The terms grant Meta a broad license to use content that businesses or their agencies post “in connection with” Meta products, authorize automated crawling and periodic pulling of associated domains and pages for development and business purposes, and retain license rights even after contracts end because of earlier July changes that made licenses perpetual. Meta issued a clarification saying the crawling language applies to advertising and business products and that robots.txt signals are still honored, but the public-facing wording had not been updated at the time of the clarification.

Consultants and security executives criticize the changes as reversing accountability, arguing that Meta builds and controls the models yet disclaims responsibility while claiming broad rights to collect and reuse enterprise content. Removing the word “train” from the wording did not blunt the effect, since collection for “developing and maintaining” products still permits extensive AI use. The combination of broad licensing, an “as-is” stance, expansive indemnity, and Meta’s ability to revise the list of covered products forces CIOs to inventory dispersed brand accounts, rethink vendor relationships, and confront stark legal and data-control tradeoffs.