Meta's new terms may allow AI training on news sites without licenses

A report from Nieman Lab indicates that Meta's upcoming terms of service may authorize the company to crawl and scrape external websites connected to businesses on its platforms, including news publishers. Under the terms, which take effect on October 30, Meta may use that outside content for purposes including developing, improving, researching, and training its AI products. The Verge, which highlighted the report, notes that the arrangement could let Meta train its AI models on publisher content without entering into a licensing agreement.

The policy covers business accounts that promote their content across Meta's platforms, including Facebook, Instagram, WhatsApp, and Messenger. Nieman Lab's analysis suggests the language extends Meta's reach to websites 'associated' with those businesses—a category that would include many news organizations. As a result, publishers that maintain a presence on Meta's apps could see their content used for AI development without direct compensation.

Nieman Lab points out that the terms appear to let Meta bypass the licensing agreements that typically govern the use of news content for AI training. Many AI developers have struck paid deals with publishers to access archives and current articles. If Meta's terms are enforced as written, it could undercut those arrangements by relying on its platform relationships instead.

The new terms take effect on October 30. The report raises questions about the scope of the policy and its impact on news organizations that rely on Meta's platforms for distribution. Nieman Lab's findings were highlighted by The Verge.