Meta drops terms clause that hinted at AI training on publisher sites

Meta has pulled a provision from its recently revised commercial terms of use after critics warned it could let the company scrape publishers' websites to train its artificial intelligence models. The removal came to light on 17 October 2026.

The updated terms had stated that Meta 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 its products and services. The wording quickly drew attention from publishers concerned about how their material might be used.

Experts saw a broad grant

Legal experts described the clause as an unusual provision. In their reading, it could have given Meta permission to crawl the site of any publisher holding a Meta business account and feed that content into the company's AI training pipelines.

A Meta spokesperson initially maintained that the revised terms did not signal any change in how the company crawls the web. The provision has since been taken out of the document.

In a statement, Meta said the crawling terms it had consolidated applied only to its advertising and business products, and that it had made this explicit in the terms themselves. "It does not change how Meta uses web crawlers," the company said.

A sensitive moment for AI and copyright

The episode unfolds while AI developers face sustained legal scrutiny over how they gather training data. Meta won a high-profile copyright infringement lawsuit last year, but other cases against AI companies, Meta among them, are still under way.

The quick reversal underlines how closely publishers now read the contracts of large technology platforms. Even a single sentence suggesting content could flow into model training is enough to trigger pushback, as media organisations push to keep control over how their work is used.