The Maze: Meta, the owner of Facebook, Instagram and WhatsApp, has removed disputed AI-training language from its proposed business terms. The change matters to companies using its services internationally: the terms still cover collecting content from associated websites. Meta says its website-reading practices have not changed.
Source: Meta. The revised terms still authorize collection from associated websites for advertising and business products.
The Details:
The wording changed before the deadline. The proposed terms take effect on October 30. An October 5 archived version allowed Meta to collect content from websites associated with a business and explicitly listed “train” among its permitted uses. Training means using examples to teach artificial intelligence models. Meta has since removed that sentence. The revised clause names advertising and business products as the purposes for collection. That narrows the stated uses, but it keeps the permission to collect website content and periodically retrieve updates. The affected material therefore extends beyond posts a company uploads to Facebook or Instagram.
A business account carries a wider relationship. These commercial terms cover activities including advertising, selling products, maintaining a product catalog and managing a business Page. They apply beyond the United States, although separate product terms can govern particular services. Under the update provision, continuing commercial use after notice means accepting the terms; a business that disagrees can stop that use. Earlier advertising-specific terms already permitted collection from website addresses supplied for ads. The proposed general terms reach sites associated with the business. Earlier terms also contemplated improving AI models, so this is not the beginning of every possible AI use.
Meta’s assurance leaves a disputed boundary. Meta says the revision clarifies existing advertising and business arrangements. It says website owners can still signal preferences through robots.txt, a file telling automated website-reading programs which pages they should avoid. The company denies gaining new rights that override those preferences. However, legal specialists remain concerned that permission to develop business products could encompass AI training. The technical standard for robots.txt also says its instructions are not access authorization. Neither the wording change nor the company’s statement independently proves which sites Meta collects from or how their content is used.
MarketMaze Take: Meta has narrowed a sentence, while leaving businesses to assess a broader relationship. The useful distinction is between the permission a platform seeks, the instructions a website publishes and what its software actually does. A change to one does not establish a change to all three.
Audience access gives the dispute commercial weight. One publisher, Stephen Beale of The Steampunk Explorer, described relying on Facebook to reach readers and send them to his website. He was considering leaving before the revision. That illustrates the tradeoff for businesses in any market: refusing content-use terms can also mean giving up a distribution channel. His experience does not establish how many companies would leave or what they would lose.
“Train” is not the whole permission. The retained ability to collect associated-site content and develop business products matters alongside Meta’s assurance about respecting preferences. The remaining legal scope is contested. Businesses have a clearer description of purpose, but no independently demonstrated end to training or court decision settling this dispute.
Sources: TechCrunch, The Verge, Meta, Meta archived terms, Meta advertising terms, Nieman Lab, Nieman Lab, IETF
