Meta Backs Off Plan to Use Business Websites for AI Training
Meta has dropped new wording that would have let it use content from business customers' websites to train its AI and improve its products.
The removed sentence said Meta could use anything found on a business account's domains, web pages, or URLs to "provide, develop, improve, research, train, and maintain the integrity of Meta products and services." TechCrunch That covered material on a company's own website, not just photos or posts uploaded to Facebook or Instagram.
The terms were set to start on October 30. A business that kept using Facebook, Instagram, or WhatsApp after that date would have automatically agreed to let Meta copy, or scrape, its websites. Nieman Lab No extra signature was needed. Staying on the apps counted as consent.
Legal experts told Nieman Lab the wording could have let Meta copy the website of any publisher with a Meta business account and use those articles to train AI models. TechCrunch That turned a routine terms update into a question about where AI gets its reading material. Any publisher with a business account would have been in scope.
Meta said the update was a consolidation of existing rules, not an expansion. A spokesperson said it did not reflect a change in how Meta crawls the web. The company said the crawling terms it had combined applied only to its ads and business products, and the clarification does not change how it uses web crawlers.
For now, the contract question is settled. The broad use language is no longer in the commercial terms.
The broader context here is about clear rules for web copying. Think of it like a parking permit that suddenly seems to cover your garage at home. When rules for ad accounts and rules for collecting AI data are mixed together without clear limits, businesses cannot tell what will happen to content from their own sites.
In my view, companies now need to keep these rules clearly separate. Business customers want simple answers: what is taken in, what it is used for, and how long it is kept. A sentence that lists product work, research and training together will be read as a claim on AI training data, even if that was not the intent.
One thing worth flagging for small teams is automatic agreement by continued use. It is efficient, and it is easy to miss. Tracking terms changes and deciding who reviews them across marketing, legal and engineering reduces surprise. The change came before October 30, which leaves time to tighten that process while keeping posting and ad setups in place.
The optimistic read here is simple. Clear limits help both sides. Publishers stay in control of their own websites. Meta keeps trust in business tools that need accurate crawling for ads and measurement. Removing unclear wording protects that exchange.
