Meta's Business Terms Could Allow AI Crawling of Websites

The permission is in the contract, not in a crawler setting

Most fights over AI training data start with a technical question: did a bot respect a block, a licence or a paywall? This one starts somewhere quieter. Meta has revised the Commercial Terms that govern business use of Facebook, Instagram, WhatsApp and Messenger. The revision, effective 30 October 2026, contains language that — as reported by Nieman Lab and The Register — authorizes automated access to domains, pages and URLs associated with a business customer's products or services, and to material made available through use of Meta's products, and allows Meta to use what it collects for a list of purposes that includes training its products. Meta's own terms page could not be fetched directly for this article; the clause wording here is as those two reports carry it.

The trigger is broad by design. Business use, under the terms, is not limited to advertisers with a contract. It reaches anyone running ads, listing products, managing a Page or using Meta Business Suite. Nieman Lab, which first reported the clause on 15 October, notes that this definition covers virtually every news organization, including itself, because nearly all of them operate at least one Meta business account. The Register followed on 16 October with additional expert comment, and the News/Media Alliance condemned the update the same day. A clause that looks like ad-tech housekeeping therefore lands on publishers, retailers, creators and any other business with a website and a Page.

Two details in the reported text deserve attention. The first is the phrase covering test credentials: the terms permit Meta to create credentials as part of its automated access. Cornell law professor James Grimmelmann told Nieman Lab that, on a broad reading, that could allow Meta to register as a user to reach content behind a registration wall. A registration wall is not a paywall, and the reporting establishes neither that Meta has done this nor that it holds an enforceable right to do it. Meta declined to answer whether such credentials would be used to reach paywalled material. The second detail is what the text does not say. It does not mention robots.txt, the standard file site owners use to tell crawlers where they may not go.

Meta's answer and the experts' answer point in different directions

Meta's response, given to Nieman Lab by email, is that nothing practical changes. The company says the new wording gathers crawling language that was previously spread across several sets of terms, including its generative AI ad terms, into one place, that its crawling of publishers will not change, and that owners can still signal their preferences through robots.txt.

The lawyers who read the document are less reassured, and their objection is specific. Robots.txt communicates a website operator's crawler-access preferences; it does not by itself establish a comprehensive legal prohibition. A set of commercial terms can create contractual permissions, but those permissions are themselves subject to applicable law and to whether a court would enforce them. The experts' position is narrower than the alarm suggests: Meta could argue that the terms authorize crawling despite a site's robots.txt instructions, and whether that argument would prevail is unresolved. It has not been tested in court. One further wrinkle, noted by Santa Clara University law professor Eric Goldman in The Register: Meta's robots.txt assurance is not expressly incorporated into the reported clause. The reassurance sits in an email to a newsroom. The permission sits in the contract.

Goldman raises a second, separate question that is not about crawling at all: how a change this broad is imposed. Customers with existing commercial relationships may have little practical ability to reject a unilateral amendment, particularly where prepaid services or other commitments are involved. That is an enforceability question about the amendment itself, distinct from whether the crawling language is broad. It is raised here as a question, not as a finding that the change is unlawful.

What a business can actually do before October 30

The practical picture, on the reported text, has three parts. A business can accept the clause and keep its Pages, ads and catalogues. It can rely on robots.txt and on Meta's statement that practices will not change, accepting that this is a policy assurance rather than a limit written into the terms. Or it can stop using Meta's products for commercial purposes before 30 October, which for most publishers and retailers means giving up a distribution and advertising channel they did not build and cannot replace quickly.

That menu is the practical picture, not legal advice, and it is not exhaustive. The reported terms offer no dedicated opt-out from the crawling clause, and Meta's notice says users who reject the updated terms should stop commercial use of its products. That is the company's stated contractual choice. It is not proof that a business has surrendered every other protection: existing agreements, copyright and applicable law may still bear on what Meta can do with collected content, whatever the terms claim.

What the record does not support, as of 16 October 2026, is a stronger claim in either direction. There is no evidence in the reporting that Meta has already crawled any site under these new terms; they are not yet in force. There is also no commitment in the terms that collected content will stay out of AI training; training is one of the uses the text lists. The News/Media Alliance has publicly called on Meta not to implement the change on 30 October; the examples in its statement, involving a performer's website media and a stock-photo library's images, are hypothetical illustrations in an advocacy statement, not documented instances of Meta collecting such material. Whether that pressure, or a legal challenge, moves the text before the deadline is the next fact to watch.

Sources

[1] Nieman Lab, Andrew Deck and Hanaa' Tameez, "Meta's new commercial terms will allow the scraping of news sites, legal experts say", 15 October 2026 (original reporting; clause text as reproduced there, Meta's emailed response, and comment from legal experts who reviewed the terms)Read source
[2] The Register, "Meta grants itself right to scrape content from business customers for model training", 16 October 2026 (clause text, business-use definition, and Eric Goldman's comment on enforceability, existing relationships and robots.txt)Read source
[3] News/Media Alliance, statement on changes to Meta Commercial Terms of Use, 16 October 2026 (industry reaction; calls for the terms not to be implemented on 30 October)Read source
[4] Meta, updated Commercial Terms (preview, effective 30 October 2026) — primary document, linked from both reports above. Direct fetch of this page returned an access error, so the clause wording in this article is as reported by Nieman Lab and The Register, not independently checked against the live documentRead source

THE ALFW VERDICT

The contract deserves more scrutiny than the reassurance. As reported by Nieman Lab and The Register, Meta's revised Commercial Terms contain broad permission to crawl websites associated with commercial users, with product training among the listed uses and no dedicated contractual opt-out. Meta says the update consolidates existing language, will not change its crawling practices and leaves robots.txt controls available — yet the reported clause does not expressly incorporate that assurance, and legal experts question both its reach and its enforceability. No crawling under the revised terms has been demonstrated, and the terms are not yet in force. The immediate issue is the breadth of the permission Meta is seeking before 30 October, not proof that it has exercised it.