Meta's October 30th terms may allow AI training on news sites

Meta has followed OpenAI in striking AI licensing deals with news organizations, but the new commercial terms described by Nieman Lab could give it access to a much broader swath of content without such deals. The terms apply to business accounts that promote content across Facebook, Instagram, WhatsApp, and Messenger, and authorize Meta to use automated software to crawl domains associated with those businesses' products or services. Legal experts tell Nieman Lab that the language could allow Meta "to argue its commercial terms supersede other blocking instructions," like a site's robots.txt file — potentially letting Meta bypass a licensing agreement and freely train its AI models on content posted by online publishers. Cornell professor James Grimmelmann tells Nieman Lab that for news publishers this clearly represents permission to train AI on their content and possibly deploy some of it in competition with the publishers. Meta's vice president of communications, Andy Stone, writes on X that the update isn't meant to target news sites, saying it's designed to pull accurate product information for ad and content recommendations. Meanwhile, Meta last month launched a Meta One subscription requiring businesses and creators to pay to include links in posts and reels, which Nieman Lab frames as Meta charging users to post links to content it will be permitted to crawl for free.