Meta drops terms language that critics said licensed AI scraping of news sites

Meta has quietly removed a passage from its commercial terms that publishers and legal scholars warned could have given the company a contractual basis to scrape news sites and train its AI on their content. According to The Verge, which reported the change on October 17, 2026, the deletion came after a wave of criticism — and after Meta representatives had spent the preceding days insisting the language never meant what critics thought.

What the removed clause said

The disputed text appeared on Meta's commercial terms page and stated that the company "may use any content from such domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of" Meta's products, services and features. The Verge notes that the wording is now gone from the page entirely.

Observers including Nieman Lab and James Grimmelmann, a professor of digital and information law at Cornell, had read that clause as an attempt to secure a free license to use scraped content — including news articles — for AI training, and potentially to override blocking instructions in a site's robots.txt file, the standard mechanism publishers use to signal which crawlers are welcome.

Meta's response

Before removing the language, Meta publicly rejected that interpretation. Andy Stone, Meta's vice president of communications, and spokesperson Daniel Roberts told critics that "there is no change to how Meta uses web crawlers; any suggestion to the contrary is false."

After the update, Roberts went further in a follow-up statement to The Verge. "We've heard you and want to clarify: the crawling terms we consolidated only applied to our ads and business products, and we've clarified that in the terms themselves," the statement said. "It does not change how Meta uses web crawlers. As always, content owners can use robots.txt to signal their preferences. This is not about scraping news sites, and these updates don't give us new rights that supersede those preferences."

Why the wording alarmed publishers

The episode highlighted a gap that matters beyond press releases. What a company's crawlers do today is one thing; what its contracts claim the right to do is another. Terms of service are documents that courts, regulators and business partners actually read, and a clause asserting the right to "use any content" for training purposes could function as leverage — or as a defence — regardless of the company's stated behaviour.

Meta's explanation, that the consolidated crawling terms were only ever meant to cover its ads and business products, cuts both ways. If that was the intent, the broad phrasing nonetheless sat in public view long enough for experts in media law to conclude that a far wider right was being claimed. The removal of the clause resolves that ambiguity, but it also underscores how easily ambiguous language could have stood unchallenged.

For publishers, the sequence is a reminder that robots.txt is a convention, not an enforcement mechanism, and that contractual language from a platform of Meta's scale can quietly reshape assumptions about how news content may be reused.

Why it matters

The reversal is a rare, concrete win for publishers in the fight over AI training data, and it shows that sustained public scrutiny of terms-of-service edits can force a company of Meta's size to retreat. But the underlying conflict is unresolved: AI developers still want access to the web's content, publishers still want payment or control, and the legal status of scraping for training remains contested in many jurisdictions. Meta dropped the language; it did not give up the goal of training models on web data, and future attempts — contractual, technical or negotiated through licensing deals — remain likely. Publishers who blocked Meta's crawlers during the dispute have, for now, kept the right to do so on paper as well as in practice.