Meta Removes Terms That Could Be Read as Allowing Use for AI Training, Says Web Crawling Practices Are Unchanged

Meta has removed a clause from its Commercial Terms that could be read as allowing it to use the contents of business account domains, webpages and URLs to provide or develop products and train AI, among other purposes.

The move comes after legal experts raised concerns that the language could authorize the collection of publishers’ and other websites’ content for AI training. Meta says the consolidated crawling terms apply only to advertising and business products, and that its use of web crawlers has not changed.

The Ripple Effects of a Single Added Sentence

The issue began with a recent update to Meta’s Commercial Terms. The terms included language stating that content on domains, webpages and URLs associated with business accounts could be used to provide, develop, improve, research and train the company’s products and services, as well as maintain their integrity.

The inclusion of the word “training” caught readers’ attention. U.S. legal experts reportedly viewed the language as a “notable clause” that could give Meta permission to scrape websites belonging to publishers with Meta business accounts and use the content to train AI models.

Copyright disputes over AI are still ongoing. Meta won a high-profile lawsuit last year in which it was accused of copyright infringement, but other lawsuits against AI companies, including Meta, remain in progress. In this climate, even a single sentence in a set of terms can be read as carrying significant weight, separate from how the company actually operates.

What the Deletion and Meta’s Explanation Show

A Meta spokesperson said from the outset that the updated terms would not change how the company crawls the web. The clause in question was later reported to have been removed.

In a statement, the company said it had clarified in the terms that the consolidated conditions on crawling apply only to advertising and business products. It added that its use of web crawlers itself has not changed.

One distinction to keep in mind is that “removing language” and “changing actual practices” are separate matters. The deletion narrowed the scope for interpreting the terms, but Meta had said from the beginning that its practices were unchanged. The company has not announced any change to its collection methods or its policy on using data for AI training.

The Clarity of the Terms Is at Issue

Even when terms are revised for consolidation or simplification, wording that readers can interpret broadly can cause concern. This case appears to show how heightened interest in AI training data amplified reactions to a single sentence in the terms.

For companies that operate their own websites and creators who publish content, it is important that the wording make clear which products the terms apply to and how far they extend. Terms in the AI era must not only be legally accurate; they also need to be written so ordinary users are less likely to misunderstand their intent.

When revising their terms, platforms should carefully explain the scope of their application, including how they relate to AI training. Businesses, too, would do well to make a habit of reviewing clauses on data handling whenever terms for advertising or business products are updated.