Meta wrote itself a very large permission slip, got caught reading it aloud, and then explained that it had never meant to use it. The company has removed a clause from its commercial terms of use that legal experts feared could let it scrape publishers' websites for AI training.
What the clause said
Meta recently updated its commercial terms. One passage declared that the company 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" Meta products and services.
The list of verbs is long and mostly routine. Then comes "train," which in 2026 is rarely a neutral word. Placed next to "any content" and "domains, web pages, or URLs," it read to some people as something much broader than ad tooling.
How the experts read it
Legal experts consulted by journalists described it as a "remarkable provision." Their concern was specific. Any publisher with a Meta business account could, under that wording, have given Meta permission to scrape its website and use the content to train the company's AI models.
Publishers are exactly the kind of organisation that holds a Meta business account. They use one to run pages and buy ads, and they rarely expect the paperwork to double as a content licence.
The background matters here. Meta won a high-profile copyright infringement lawsuit last year over how it trained its models. But other cases against AI companies, Meta among them, are still going. In that climate, a clause that might simplify the consent question is bound to get read closely.
Meta's answer
A Meta spokesperson insisted from the start that the updated terms did not reflect any change in how Meta crawls the web. The provision has now been removed altogether.
The company's statement said: "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." It added: "It does not change how Meta uses web crawlers."
Two details in that statement deserve attention:
- "Consolidated." Meta describes the change as housekeeping, the merging of existing crawling terms into one place, not a new grab for content.
- "Clarified in the terms themselves." The fix was to the document, not just to the company's description of it. That is a quiet admission that the original text left room for the broader reading.
Where this leaves publishers
The reported facts support two readings, and they do not fully reconcile.
- The experts' view. The wording could be read as permission to scrape any business account holder's site and train AI on it.
- Meta's view. The language covered only ads and business products, and nothing about its web crawlers changed.
The removal ends the immediate worry, since the clause is no longer in the terms. It does not settle what the original text meant, and Meta's explanation is the company's own account of its intent. What the episode does show is how fast a single sentence in a commercial agreement can turn into an AI training question.
For anyone who manages a business account, the practical lesson is dull but real. Terms of service are no longer just about ad billing. When a company that trains AI models updates them, someone should read the new version.
Meta, for its part, has resolved the matter in the most efficient way available. If a sentence causes alarm and was never supposed to mean that, you can delete the sentence.