Meta has removed a contested clause from its new commercial terms of use that would have let it use content from business customers' linked websites to 'train' its products, after publishers and legal experts warned the wording could open the door to scraping news sites for artificial intelligence.

The terms, due to take effect on 30 October, cover businesses that use Meta's advertising, Pages, apps and Business Suite. An earlier version said Meta could use automated software to crawl the domains, web pages and URLs 'associated with' those accounts and use any content found there to develop, improve, research and train its products.

Legal experts who reviewed the draft for the Nieman Journalism Lab said it could be read as permission for Meta to scrape publishers' sites and train AI, and possibly to override the blocking instructions that sites set in their robots.txt files. The News/Media Alliance called the approach an abuse of market power.

Meta says its practice will not change. '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,' a company spokesperson, Daniel Roberts, said. Content owners can still signal their preferences through the robots.txt file, he said.

The company's vice-president of communications, Andy Stone, said the update was designed to pull accurate information about products sold on third-party websites for advertising and content recommendations, and was not aimed at scraping news sites.

Meta consolidated several older policies into the commercial terms, which businesses must accept to keep using its advertising and marketing tools. Publishers have been pressing technology companies to pay licences for AI training on their content.