Meta Deletes AI Training Clause After Publisher Backlash

By Daniel Okafor

Meta has removed a clause from its commercial terms of use that legal experts said could have licensed the company to scrape publishers' websites and train its AI models on the content. The Verge first reported the removal. The clause had lived in Meta's terms for only a short stretch before the walk-back.

The disputed language appeared in an update to Meta's commercial terms. It stated that Meta can "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.

That wording set off alarms among legal experts interviewed by Nieman Lab. They called it a "remarkable provision." Read literally, it gave Meta a contractual hook into any publisher running a Meta business account. The publisher's own pages, under its own domain, would become raw material for Meta's AI training pipeline. No negotiation. No separate consent.

Why did publishers see a legal trap?

The mechanism was simple. Any news outlet or brand that uses Meta's business tools — the account required for advertising, page management and commerce — would have agreed to the terms. The clause then stretched that agreement from Meta's platform to the business's entire web presence.

Context matters here. Meta defeated a high-profile lawsuit over alleged copyright infringement last year. But other cases against AI companies, including Meta, are still working through the courts. A contract term like this one would have given Meta a second line of defense: not fair use, but consent. Publishers would have signed it themselves, buried in boilerplate.

A Meta spokesperson had insisted, even before the deletion, that the updated terms did not reflect a change in how Meta crawls the web. The company's argument was that the clause merely consolidated existing language, not that it expanded Meta's rights.

What did Meta actually say?

After the removal, Meta issued a statement that addressed the criticism head-on.

"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 company said. "It does not change how Meta uses web crawlers."

That is a rare public reversal. Meta did not defend the clause on the merits. It removed it and framed the episode as a drafting problem — a consolidation of crawling terms that readers misread as a rights grab.

The gap between the text and the intent remains the story. The clause said Meta could use business-account content to "train" its products. Meta now says that language applied only to ads and business products. Publishers had no way to know that scope from the words on the page. They learned it only after the backlash forced an explanation.

What does the episode change?

Not much in Meta's crawling behavior, if the company's statement holds. Meta says its web crawlers operate today exactly as they did before the terms update. The removal of the clause restores the status quo on paper.

But the episode signals something broader. AI training rights have become the most contested clause in platform terms of service. Publishers now read every revision of Meta's contracts the way they read regulatory filings — line by line, looking for a single word like "train." Nieman Lab's experts flagged this provision within days. That scrutiny is now permanent.

The episode also shows the leverage that public pressure still carries. Meta absorbed a copyright ruling in its favor last year and faces ongoing litigation. Yet a wave of expert criticism over one clause pushed the company to strike it within days, rather than defend it in court later. For any business that runs a Meta account, the practical lesson is to audit platform terms after every update — because the next consolidation may not be walked back so quickly.