Meta Drops AI Training Language From Business Terms Publishers Said Let It Crawl Their Sites

Meta has removed the word "train" from a website crawling clause in its revised Meta Commercial Terms, after news publishers and legal experts said the clause would let it scrape their sites for AI training. The company changed the terms late on October 16, 2026, Pacific time, and said the crawling rights only ever applied to its ads and business products.

The terms take effect on October 30, 2026 and cover any business use of Meta's apps, from running ads to managing a Facebook Page. Nieman Lab first reported the wording on October 15, and it touched a sore point in the business of AI: whether AI companies pay for the news they use. Meta announced AI licensing deals with CNN, Fox News, USA Today and People Inc. on December 5, 2025, so its Meta AI chatbot could answer with their reporting (The Verge).

What did Meta's original crawling clause say?

The first version of the clause authorized Meta to use automated software and create test credentials to crawl any domain associated with a business account, and to use that content to "train" its products, according to The Verge, which quoted the clause in full. The listed tools run from web scrapers and bots to spiders, crawlers and user agents.

Because managing a Page or running ads falls under the terms, nearly every news organization has a business account that would be covered, Nieman Lab reported. James Grimmelmann, a professor of digital and information law at Cornell, told Nieman Lab the language could let Meta train AI on publishers' content and possibly deploy it in competition with them. Legal experts also said Meta could argue the contract overrides robots.txt, the plain text file sites use to tell crawlers which pages they may visit.

Meta first pushed back. Spokesperson Daniel Roberts said nothing had changed in how Meta uses web crawlers, and Andy Stone, Meta's vice president of communications, wrote on X on October 16 that the update was about pulling accurate product information from other sites for ad and content recommendations, not about news sites (The Verge).

What changed in the revised terms?

The revised clause, as it reads on Meta's terms preview on October 17, 2026, limits the purpose of crawling to advertising and business products and to the integrity and security of Meta's products, and no longer contains the word "train". Meta also folded the separate sentence on how it may use crawled content into the authorization itself.

Roberts added that content owners can still use robots.txt and that the update gives Meta no new rights that override those preferences, The Verge reports.

Was the training language new for Meta's advertisers?

Not entirely: Meta's current Self-Serve Ad Terms, which carry an effective date of August 22, 2025, already let Meta crawl the sites of anyone who places an ad order for uses that include "research and training." The current Commercial Terms, dated March 6, 2024, contain no crawling clause.

The older ad clause is narrower than the first Commercial Terms draft in two ways: it applies only to businesses that place an ad order, and it says nothing about test credentials. The revised Self-Serve Ad Terms, also due on October 30, replace it with the same wording as the revised Commercial Terms, without training.

Meta told Nieman Lab the crawling language was previously spread across other terms, and the ad terms bear that out in part. They also show what publishers reacted to: in the Self-Serve Ad Terms the clause applies to ad orders, while in the Commercial Terms it covers every business use, including managing a Page or using Meta Business Suite. On training, the October 30 wording is narrower than what advertisers have accepted since August 2025.

Why are publishers still objecting?

The News/Media Alliance, a US publishers' trade group, said the revision limits what Meta may use content for but not its ability to crawl and collect it. Spokesperson Sam Quigley told the Mercury News that training was taken out of the listed uses without being ruled out, so "it's still potentially on the table."

Eric Goldman, a law professor at Santa Clara University, told the same paper the language is narrower but "still vague and potentially overbroad," pointing to the catchall on integrity and security. He also asked why Meta does not write its robots.txt commitment into the terms.

The test credentials remain as well. Grimmelmann told Nieman Lab they could let Meta sign up as a user to get past registration walls, and Meta declined to say whether they would get past paywalls. The revised clause still authorizes Meta to create test credentials, according to the terms preview.

What we don't know yet

• Whether Meta's crawlers will treat robots.txt rules on business accounts' sites as binding after October 30, 2026.
• Whether content crawled under the August 2025 ad terms wording on research and training has been used to train Meta's AI models.
• What maintaining the integrity and security of Meta's products covers, and whether it could include model training.
• How Meta will use test credentials on sites with registration walls or paywalls.