← All articles

Mikhail Polyakov

Meta's new terms for business accounts: what they say about crawling websites and training, what Meta replies and what to check before 30 October

Level: BeginnerReading time: 6 min

Topics: AI, Small business

A thick contract, glasses and a pen on a shop table, with a small brass mechanical spider crawling towards a stack of product cards

The short answer: on 30 October 2026, Meta's updated commercial terms for business accounts come into force. Under them, a company authorises Meta to crawl with bots the websites associated with its products or services, to create test credentials for this, and to use the content of those websites, including to "train" Meta's products. According to Ars Technica, it is not possible to opt out of this clause: the alternative is to stop using Meta's services. Meta says it has merely gathered scattered wording in one place and that the way its bots work is not changing. Below: what the terms say, what the dispute is about and what a company should check.

Where the news comes from

I read the Ars Technica article (author Ashley Belanger, 16 October 2026). It relies on reporting by Nieman Lab, on its own comparison of the old and new terms and on replies from a Meta spokesperson. I did not open the text of the terms on the Facebook website or the Nieman Lab piece myself: everything said about them below is as reported by Ars Technica.

What the new terms say

According to the quotation given by Ars Technica, a company with a business account authorises Meta, or anyone acting on its behalf:

  • to use automated software (bots, web crawlers) to access domains, pages and URLs associated with its products or services, or made available through the use of Meta's products, to index, store, analyse and collect their content and to periodically pull updates;
  • to create "test credentials" for this. What exactly is meant by them is not spelled out in the quoted text; going by the words, these are accounts that allow a program to log in to a website. The experts quoted by Nieman Lab fear they could be used to bypass paywalls;
  • to use any content from these websites to provide, develop, improve, research, "train" and maintain Meta's products and services.

The word "AI" does not appear in this clause, as Ars Technica notes: the reading that it is about training AI rests on a single use of the word "train" and on the references to automated crawling of websites.

Why lawyers are alarmed

  • New wording. Ars Technica compared the current terms with the new ones. The current commercial terms contain no "test credentials". According to the experts quoted by Nieman Lab, these could allow paywalls to be bypassed and materials to be collected automatically. Meta pointed the publication to its seller agreement, where such credentials were already mentioned, but there, as Ars Technica writes, the purpose is narrow: to improve the seller's product listings.
  • A wider purpose than before. Previously, according to Ars Technica, crawling was mentioned only in the terms on generative AI in advertising, and only for the business's own ads. The new clause covers all of Meta's products.
  • The lawyers' assessment. Three lawyers told Nieman Lab that it is unclear whether a court would enforce such a clause, but that it could be used to argue that Meta may collect content from news websites and train its AI products on it. Cornell Law School professor James Grimmelmann believes Meta could train AI on any content from those websites and possibly show some of it on its own platforms in competition with the publishers themselves. Lawyer Matt Topic called the clause remarkable because access to such material on a licensing market would cost many millions of dollars.

This affects news publishers most: according to Nieman Lab, almost every one of them has at least one Meta business account. But the terms apply to all business accounts.

What Meta replies

  • In a blog post, Meta called the update "streamlining": wording on crawling used to be scattered across different terms, and now it has been gathered in one place.
  • A Meta spokesperson told Ars Technica that nothing is changing in how Meta uses web crawlers, and that any suggestion to the contrary is false.
  • A Meta spokesperson told Nieman Lab that Meta has no plans to use these terms as a basis for ignoring robots.txt, the file in which a website forbids crawling. Content owners, according to Meta, can still manage access through robots.txt, and Meta provides instructions on how to block crawling.

News publishers, according to Ars Technica, question whether this position would stay the same if courts found AI training on articles that are not news to be permissible (this is how Ars Technica words the condition).

What here is fact and what is interpretation

This is my own breakdown, not words from the source.

  1. Fact from the text of the terms. The right to automatically crawl linked websites, to create test credentials and to use the content, including for "training"; the date of 30 October.
  2. Meta's position. This is streamlining, the way the bots work is not changing, robots.txt is respected.
  3. The lawyers' interpretation. The clause could become an argument in favour of collecting material to train AI; whether a court would enforce it is unclear.

What a company with a Meta business account should check

This is my own conclusion from the article, not advice from its author.

  1. Who does it. The owner of the company or a lawyer, together with whoever runs the website.
  2. What to read. Meta's new commercial terms in full, before 30 October, and compare them with what you are prepared to allow. A news summary, including this one, does not replace the text of the terms.
  3. What to check on the website. Whether the website has a robots.txt file and what it says about crawlers. If you do not want the website's content collected, the prohibition has to be written there explicitly. How to confirm: open the address yoursite/robots.txt in a browser and see what it shows; look for Meta's instructions on blocking crawling in its help pages. Such a check shows that the prohibition is written down, but not that it is respected: Meta promises to respect robots.txt in its replies to publications, and this cannot be verified from the article.
  4. If you have a paid or restricted section. Ask your developer whether it can be reached without paying or logging in, for example through a direct link. The experts quoted by Nieman Lab fear that test credentials could allow paywalls to be bypassed, so it is useful to know exactly how your website grants that access.
  5. If the terms are unacceptable. According to Ars Technica, it is not possible to opt out of the clause; the alternative is to stop using Meta's services. Weigh how many customers Facebook and Instagram bring you before deciding.

How to read news like this. Tell the text of the terms apart from its interpretation. The text is quoted here, but what it means in practice will be decided by the courts and by Meta's own actions, not by headlines.

Summary

From 30 October 2026, Meta business accounts agree to automated crawling of the websites linked to them and to the use of their content, including for "training" Meta's products. Meta calls this streamlining and promises to respect robots.txt; lawyers see the clause as a possible argument for collecting material for AI. A company with such an account should read the terms before that date and check what its own website tells the bots.

See also another piece in this section: USA Today sues OpenAI over its articles.

I work on AI agents and automation. If you would like to see my projects or discuss your own task, take a look at my portfolio.

Sources