Meta has quietly walked back one of the broadest AI-training claims any platform has put in writing. The company removed language from its commercial terms that said it could 'use any content from [a business account's] domains, web pages, or URLs to provide, develop, improve, research, train, and maintain' its products and services, TechCrunch reported on Saturday.
Legal experts told Nieman Lab the provision looked like permission to scrape the website of any publisher holding a Meta business account and feed the content into Meta's AI models — a reading that triggered immediate backlash from publishers already in litigation over AI training data.
After the backlash, Meta removed the provision. A spokesperson said the terms 'did not reflect a change in the way Meta crawls the web' and that the consolidated crawling terms 'only applied to our ads and business products.'
The episode lands in a charged legal landscape: Meta previously defeated a high-profile copyright lawsuit over alleged AI-training infringement, but other cases against AI companies, including Meta, continue through the courts. For publishers, the removal closes this clause — but not the underlying fight over who consents to what when a platform's terms become a training licence.
