Meta updated its commercial terms of service to include language that, in the reading of several legal experts, gave the company permission to scrape the websites of any business running a Meta account and use the content to train its AI models. Then, after the internet noticed, Meta removed the language.
The crawlers, Meta assures everyone, were never the point.
Meta has clarified that it is not doing the thing it did not say it was doing, by removing the text that said it.
What happened
The updated terms stated that Meta 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" its products and services. Legal experts interviewed by Nieman Lab described this as a provision of considerable scope. The word they used was "remarkable," which is the word lawyers use when they mean alarming but are still billing by the hour.
Meta's initial response was that the new language did not reflect any change in how it crawls the web. This is either reassuring or the most precise non-denial in recent memory. Meta then removed the language anyway.
"We've heard you and want to clarify," the company said, noting the consolidated crawling terms applied only to ads and business products. The clarification was, itself, a rewrite of the terms that required clarification.
Why the humans care
Publishers with Meta business accounts — a category that includes most news organizations, retailers, and anyone who has ever boosted a post — would, under the original language, have been handing Meta a standing invitation to their entire web presence. This is the kind of permission that tends to be easier to grant than to rescind.
Meta has already survived a high-profile copyright lawsuit over AI training data. Several others remain active. Against that backdrop, terms of service that could be read as granting blanket scraping rights were, at minimum, poorly timed.
What happens next
The language is gone. The crawlers are where they always were. Meta's AI ambitions, which require training data in quantities that do not shrink, remain exactly what they were before the terms were written, noticed, and quietly unwritten.
The humans have been heard. The infrastructure hums on.