Meta pulls AI training clause from commercial terms after scraping backlash

What happened

Meta has stripped a contested clause out of the commercial terms that govern its business tools, retracting language that legal experts read as permission to scrape publishers' websites for AI training. The Verge reported the removal, which came after the updated wording began attracting criticism.

The clause appeared in a recent revision of Meta's commercial terms of use. It gave Meta the right to use content associated with a business account's domains, web pages and URLs for a long list of purposes — developing and improving its products, conducting research, keeping its services secure and, notably, training its models. That final verb is what set off alarms.

How experts read it

According to legal experts interviewed by Nieman Lab, the provision looked unusually broad — wide enough, on its face, to let Meta crawl the website of any publisher holding a Meta business account and fold that material into its AI training pipeline. For news organizations and other companies that depend on Meta's advertising and business tools, that reading suggested a substantial new rights grab buried in boilerplate.

The timing sharpened the concern. As TechCrunch reports, Meta prevailed in a prominent copyright lawsuit last year, but a broader wave of litigation over AI training data is still moving through the courts, with cases pending against several AI developers, Meta included.

Meta's walk-back

Even before the reversal, Meta rejected that interpretation. A company spokesperson insisted the refreshed terms did not change how Meta crawls the web.

The removal then followed, and Meta explained it in a statement. The company said it had heard the feedback and wanted to clarify matters: the crawling terms it had consolidated applied only to its ads and business products, and Meta had now rewritten the terms to state that explicitly. The company added that nothing about how its web crawlers operate has changed.

Why it matters

The swift reversal is less a resolution than a reminder of how much now rides on terms-of-service fine print. Businesses that use Meta's commercial tools are bound by documents they rarely read line by line, and a short list of verbs can redefine what the company may do with their content.

It also highlights the gap between what a policy says and what a company does. Meta's position is that its crawler behavior never changed, which cuts both ways: the new clause did not expand scraping, but its removal does not obviously restrict it either. Whether training AI on scraped web content is lawful at all remains unsettled, and that question is being decided in courtrooms rather than in terms of service.

For publishers, the practical lesson is vigilance. The change was caught because commentators went looking for it; the next revision may not attract the same scrutiny, and Meta has shown it will consolidate and rewrite its crawling terms when it chooses.