Meta Quietly Slipped AI-Training Rights Into Its Terms — Then Deleted Them After Backlash

 

Meta tried to rewrite the rules in its own favor this week — and got caught. The company quietly inserted a clause into its commercial terms of use that, legal experts say, would have let it scrape the websites of any business account holder and use that content to train its AI models. After days of backlash, Meta deleted the language on Saturday, October 17. Here's what happened, and why the rollback matters less than the attempt.

This isn't the first time AI training has collided with someone else's content — OpenAI recently paused training of its most powerful models over safety concerns earlier this fall.

What Meta changed

Meta added a line to its commercial terms page declaring it could "use any content from [a business account's] domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of" Meta products and services. Tucked into the terms page and set to take effect October 30, it read like a license to turn the entire business web into AI training data.

Why legal experts freaked out

Legal experts interviewed by Nieman Lab called it a "remarkable provision." James Grimmelmann, a Cornell professor of digital and information law, told The Verge it "clearly represents permission to Meta to train AI on any content from their sites." The scary part: it applied to any publisher with a Meta business account — even a local shop that just runs a Facebook page. There was no mention of robots.txt, the standard signal sites use to tell crawlers to back off, and some experts said the clause could have overridden those instructions entirely.

Meta's defense

Meta communications VP Andy Stone pushed back on X, saying the update was meant to pull accurate product listings for ad recommendations — "not about scraping news sites." As The Verge's reporting amplified the backlash, Meta removed the clause outright on October 17. Spokesperson Daniel Roberts said the terms "only applied to our ads and business products" and that "it does not change how Meta uses web crawlers." TechCrunch confirmed the removal the same day.

Why it still matters

Three things. First, the terms still take effect October 30, and Meta still authorizes itself to crawl business-account domains — it removed the word "train," not the crawling. Second, per Nieman Lab (as relayed by The Verge), Meta already pays CNN, Fox News, USA Today, and People Inc. for AI training content — the deleted clause would have let it take the same kind of content from everyone else for free. Third, it fits a pattern of platforms quietly renegotiating what they can take from your website. Anthropic just pitched Congress its own AI safety bill — accountability is coming from every direction now.

The takeaway

The deletion is good news, but the playbook is the story. These clauses land in terms pages on a Friday; the rollback only happens if someone notices. For publishers and small businesses, the lesson is boring but real: read the fine print on every platform you depend on. Microsoft's own CEO says you can't trust AI without an emergency brake — the same skepticism applies to the terms behind the AI.