Meta drops 'train' from scraping terms, but broad crawling rights remain

What's the deal? Meta has removed the word "train" from its upcoming commercial terms of use after legal experts and news publishers warned the language could let it scrape their websites to train AI models. The revised terms, published Friday, now limit crawled content to "advertising and business products" rather than all Meta products.

Why now? The change followed a Nieman LabDealroom has a profile for this one. Try Dealroom → report that the terms would permit broad scraping of any website "associated with" a business account, with no opt-out short of quitting Meta's business tools. The News/Media AllianceDealroom has a profile for this one. Try Dealroom →, a trade group representing publishers, called the original terms "outrageous and probably illegal."

What changes in practice? Less than critics hoped. The broad crawling permission stays, and the remaining word "develop" is wide enough to still cover AI training, legal experts say. The terms take effect October 30, 2026, with no option for business accounts to opt out beforehand.

Meta said the revision was a clarification, not a reversal. "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. It does not change how Meta uses web crawlers," the company said.

What could go wrong? Meta says site owners can still block its crawler via robots.txt flags, but it has not written that guarantee into the terms. "It appears as if the new language just limits what Meta is allowed to use publisher content for, but not their ability to crawl and get the content in the first place," said Sam Quigley of the News/Media Alliance. Santa Clara University professor Eric Goldman called the terms "still vague and potentially overbroad."

The signal: Publishers are increasingly trying to separate AI-training access from ordinary search crawling, with some pursuing licensing deals rather than relying on open-web scraping. Meta, which has struck deals with major publishers and already won a US lawsuit over training AI on copyrighted books, is testing how far broad crawling rights can stretch — and the lack of a clear opt-out remains the unresolved issue.

Read more: Nieman Lab · TechCrunch · News/Media Alliance