Facebook and Instagram advertisers: from October 30, Meta can train its AI on your website

Do you have a Facebook page for your shop, a business Instagram account or ads with Meta? On October 30, new business terms will take effect. The previous ones dated from March 6, 2024. The new version contains a paragraph that is drawing a reaction from news publishers.

According to the text quoted by the trade site PPC Land, a company that accepts these terms authorizes Meta, “or any entity acting on its behalf,” to use automated software on the domains and pages “associated” with its products or services. Meta may come back regularly to retrieve updates, and even create test accounts to access them. Finally, it may use this content to “provide, develop, improve, research, train” its products, including its artificial intelligence tools.

Why the press is worried

The site Nieman Lab, from Harvard University's Nieman Foundation, raised the alarm on October 15. Three legal experts consulted by the outlet believe the text could allow Meta to argue that it has the right to scrape news sites that have a business page, in order to train its AI. According to these experts, Meta could even argue that this contract overrides the robots.txt file, which is how a site tells bots what they must not crawl. The News/Media Alliance, which represents American publishers, called these terms “outrageous and likely illegal,” through its president Danielle Coffey in a statement on October 16.

There is no checkbox to refuse this one paragraph. Continuing to use Meta's business products counts as acceptance. The only way to say no is to stop using them.

What Meta says

Meta disputes this reading. Its spokesperson Andy Stone called reports that the text would allow scraping of news sites “false.” According to him, the aim is to retrieve accurate information about products sold on third-party sites. The company also told Nieman Lab that these clauses were previously scattered across several documents and that it would not change the way it crawls publishers' sites. According to Meta, site owners can still control access to their content with robots.txt.

The text itself sets no limit to product pages and does not mention robots.txt. It does, however, include a clause allowing Meta to sue non-U.S. companies in its own courts if they try to abuse its products, disrupt their operation or use them in an unauthorized way.

What you can do before October 30

• Read the new version of the business terms, notified in Meta's business tools, and list the sites you have linked to them (shop, catalog, company site).
• Check your robots.txt file and the rules of your host or content delivery network (CDN) that apply to AI bots. Meta says it respects them, but the text of the terms does not mention robots.txt.
• In Europe, the 2019 copyright directive lets rights holders object to the mining of their content through a machine-readable notice. An explicit opt-out in your legal notice and in robots.txt therefore remains worthwhile.
• If you are a publisher, ask a lawyer: which text prevails, the accepted contract or the site's opt-out? Nobody knows yet.

Meta, which has just launched its paid model Muse Wonder, needs fresh data. It remains to be seen whether the text will be revised before October 30, and whether a European authority will take it up.