Meta’s updated Commercial Terms, due to take effect on 30 October 2026, no longer contain the sentence that legal experts said would let Meta train AI on content crawled from business sites. Archived copies of Meta’s own updated-terms page show that sentence present on 5 October and still present at 16:20 UTC on 16 October, and gone from captures taken at 03:19 UTC on 17 October and 20:15 UTC on 18 October.
The authorisation to crawl those domains was not deleted with it. In the later captures the crawl clause survives with a narrower stated purpose, limited to advertising and business products and to the integrity and security of Meta’s own products, and the words “create test credentials” remain.
What the archived text says before and after
The quotations in this section come from Wayback Machine copies of facebook.com/legal/commercial_terms_preview, the page Meta used to publish the updated text. Its footer reads “Effective Date: October 30, 2026”.
The criticised sentence was present in the captures from 5 October at 01:46:40 UTC, 16 October at 11:48:51 UTC and 16 October at 16:20:00 UTC, under the heading “Licenses & Rights You Provide Meta”:
“Meta may use any content from such domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of, the Meta Company Products, services, and features provided under these Commercial Terms and any other applicable supplemental terms.”
The captures from 17 October at 03:19:35 UTC and 18 October at 20:15:50 UTC contain no version of that sentence. The crawl clause above it survives, with a purpose limit added where the earlier text simply ended at “and to periodically pull updates from such domains, web pages, or URLs”:
“You authorize us (or any entity acting on our behalf) to use automated software (such as web scrapers, bots, robots, spiders, crawlers, and user agents) and create test credentials to access, index, cache, analyze, collect, retrieve, or crawl the domains, web pages, or URLs … in order to provide, develop, improve, research, and maintain advertising and business products … and maintain the integrity and security of Meta Company Products.”
One limit sits beside this evidence. A capture proves what the page served at that moment, not the minute an editor changed it, so the edit window runs from after 16 October 16:20 UTC to before 17 October 03:19 UTC.
The wording was already public on 5 October, before the criticism
The sentence was already on the preview page on 5 October, so it was not inserted on 16 October when reporting began. Nieman Lab published three legal experts’ readings on 15 October; Ars Technica followed on 16 October, arguing that Meta’s description of the exercise as “streamlining” understated what had changed. Meta’s edit was being served by 03:19 UTC on 17 October, some twelve hours before The Verge reported the removal that day.
What was removed and what the surviving clause still authorises
The word “train” is gone, along with the reference to “services, and features provided under these Commercial Terms and any other applicable supplemental terms”; the licence clause in the same section now stops at “including to provide, develop, improve, research, and maintain Meta Company Products”. What remains is a permission to crawl, limited to advertising and business products and to the integrity and security of Meta’s products.
Meta’s terms do not define “test credentials”, which the clause lets Meta or an entity acting for it create in order to reach domains tied to a business’s products or services. Cornell Law School professor James Grimmelmann told Nieman Lab the phrase would technically let Meta “sign up as a fake user” past registration walls, and that Meta did not answer the outlet’s questions about whether such credentials would bypass paywalls.
Which document binds whom
The Commercial Terms are the business document. They apply to use of Meta’s products “for a business or commercial purpose”, which the text defines to include buying ads, selling products, managing a Page or using business tools such as Ads Manager or Meta Business Suite. Ordinary consumer use falls under the Meta Terms of Service and the Instagram Terms of Use, separate documents, and the Commercial Terms incorporate supplemental agreements by reference, among them the Self-Serve Ad Terms and the Seller Agreement.
The contracting entity depends on where the business is based. The Self-Serve Ad Terms state that Meta Platforms, Inc. provides the interfaces to businesses in the United States and Canada, and Meta Platforms Ireland Limited provides them elsewhere. The Commercial Terms in force today carry the effective date 6 March 2024, and the criticised wording never appeared in that document.
Continued commercial use after notice of an update is how the text binds a business, and the only step it names to avoid the new terms is to “stop all access or use of our Products for business or commercial purposes”. None of this is legal advice.
Did equivalent crawl rights already exist?
In narrower form, yes. The Seller Agreement, last updated 26 May 2026, allowed Meta to use automated software and “create test credentials” to retrieve information from URLs a seller provides, limited to Meta’s commerce features and the enforcement of its commerce policies. The Ad Creative Generative AI Terms, effective 6 May 2024, grant a right to “access, index, cache, analyze, or crawl” domains and URLs an advertiser provides, for that advertiser’s own ads; the wording there is “create and test technologies”, and “test credentials” does not appear.
Those documents support Meta’s account that crawl language sat scattered across several agreements. They also mark where the criticised version went wider: it covered any business or commercial user of Meta’s products, not only sellers or self-serve advertisers, and it licensed crawled content to train products and features and anything reached through “any other applicable supplemental terms”.
What Meta says, and what the experts argued
Meta’s statements are its own claims. On 15 October a spokesperson told Nieman Lab that crawl language “was previously scattered across different terms of use, including terms about using generative AI in ads, and this update consolidates them in one place”. After the deletion, Meta spokesperson Daniel Roberts said in a statement carried by The Verge and TechCrunch that “the crawling terms we consolidated only applied to our ads and business products, and we’ve clarified that in the terms themselves”, and that the updates “don’t give us new rights that supersede those preferences”.
Part of that account matches the diff, since the survivor is limited to advertising and business products. The diff also records the deletion of a sentence whose purposes included training, which the consolidation statements do not mention.
The legal readings came before the deletion. Matt Topic of Loevy & Loevy called the provision “remarkable” and said it gave Meta access to material that “would likely cost many millions of dollars in a licensing market”, while calling enforceability unclear. Benjamin Sobel of the University of Wisconsin described it as resembling “tying” in antitrust law, since using Meta’s commercial products means accepting the crawling permissions with them. James Grimmelmann read the wording as permission to train AI on publishers’ content and possibly to use some of it against them.
No source in the coverage sets out a mechanism by which a licensing payment would be avoided, and no saving figure has been published. Topic’s “many millions” is his estimate of the material’s value in a licensing market, and Ars Technica’s framing that the terms could let Meta avoid licensing costs is that publication’s assessment, which Meta denies.
What to check before 30 October
The Commercial Terms that bind businesses today carry the effective date 6 March 2024 and contain no crawling clause. The updated text, with the narrowed clause and no “train”, takes effect on 30 October 2026, the same date as the updated Self-Serve Ad Terms. Anyone wanting to see whether the surviving clause changes again can compare the live pages with the captures quoted above; Meta’s legal pages returned no readable text to our fetches on 19 October, so the archive is the checkable record.
