Meta Terms Grant Permission to Train AI on Publisher Content

Meta commercial terms update October 2024

Meta has rolled out updated commercial terms kicking in on October 30, and they look like they hand the company sweeping permission to crawl and scrape sites tied to businesses using its platforms. Nieman Lab reports the new wording targets online publishers whose domains are considered "associated" with a business account on Facebook, Instagram, WhatsApp, or Messenger. The terms explicitly say Meta can deploy automated tools-scrapers, bots, crawlers, you name it-to access, index, cache, analyze, and pull content from those domains, then turn around and use that material to build, refine, research, and train its AI products.

This applies to any business pushing content across Meta’s app ecosystem. By clicking "agree," a business effectively authorizes Meta-or anyone acting on its behalf-to spin up test credentials and periodically fetch updates from the relevant pages. Meta spokesperson Daniel Roberts frames this as a consolidation of existing commercial and self-serve ad terms, insisting nothing changes regarding how their web crawlers actually operate. He points out content owners can still signal preferences via robots.txt flags, per their developer docs.

Legal experts challenge Meta AI training permission

Legal experts Nieman Lab spoke with aren't buying that characterization. They argue the revised terms could let Meta claim its commercial agreement overrides other blocking instructions, including a site’s robots.txt file. If that holds up in court, the company could sidestep licensing deals entirely and freely train models on publisher content without cutting a check. Meta has already inked AI licensing deals with outlets like CNN, Fox News, USA Today, and People Inc., but this language potentially unlocks a much wider library of material.

James Grimmelmann, a digital and information law professor at Cornell, told Nieman Lab the terms look like clear permission for Meta to train AI on any publisher site content-and possibly deploy it in ways that compete directly with those publishers. Andy Stone, Meta’s VP of communications, pushed back on X, saying the update is strictly about ensuring accurate product info from third-party sites for on-platform ad and content recommendations, not scraping news outlets.

Businesses risk losing Meta access over terms

The leverage here is real: businesses that refuse the new policies risk losing all access to Meta’s products for commercial use. Just last month, the company launched a Meta One subscription charging businesses and creators for the privilege of including links in posts and Reels. Nieman Lab noted the irony-Meta is now charging users to post links to the very content it just gave itself permission to crawl for free.

Meta clarifies crawling terms scope ads business products

On October 17, Meta issued a clarification stating the consolidated crawling terms only apply to its ads and business products, claiming the language was updated to reflect that narrower scope. Crucially, though, they didn't yank the broader authorization language, leaving the door open for debate on its actual reach.

For anyone building with or using AI, this signals a shift in how major platforms assert control over web-scale training data. If these terms survive scrutiny, publishers and content owners could see their work absorbed into foundation models without a separate negotiation, fundamentally changing the economics of data licensing. Developers relying on open-web corpora should watch closely to see if Meta’s crawlers start indexing previously restricted sites and whether robots.txt directives actually hold water in practice. Publishers and businesses with Meta accounts need to comb through the updated terms, figure out if their sites fall under that "associated" definition, and decide whether to tweak their robots.txt settings or pursue explicit licensing deals to protect their interests.

Source: The Verge