Meta's new terms and AI trained on news sites
From 30 October Meta's commercial terms will allow it to use websites linked to business accounts to train its AI. Publishers fear losing any compensation; Meta says blocking crawlers still works.
by Mattia Panatta
Meta has updated the commercial terms that apply to anyone using Facebook, Instagram, WhatsApp and Messenger as a business. From 30 October 2026 the text will allow Meta to analyse websites “associated” with business accounts and use their content to train its artificial intelligence products.
Why publishers are worried
Almost every newspaper has a Facebook page and an Instagram profile linked to its website. According to several legal experts interviewed by The Verge, the new terms amount to permission to use every article for training, with no payment: the text contains no mention of fees or licences. And anyone who rejects the new terms would have to stop using Meta’s platforms for commercial purposes, a hard choice for any outlet.
There’s also a technical and legal doubt: many sites block AI crawlers with a robots.txt file. Some lawyers fear Meta could argue that the commercial contract overrides that block.
Meta’s response
Meta says the update merely consolidates existing wording, that robots.txt controls remain valid and that news organisations can still opt out of content collection. It has not, however, changed the text.
My take
It’s the same underlying question I raise in my opinion on AI slop: if models feed on writers’ work without paying for it, who will keep producing quality content? In Europe the issue intersects with the AI Act and with copyright law, which lets authors reserve the use of their texts for training.
If you have a business website linked to a Facebook or Instagram page, the practical advice is simple: check your robots.txt file and, if you don’t want your content used for training, state it explicitly in your site’s terms too.
