Meta removes terms language that legal experts read as permission to scrape sites for AI training — and insists its crawlers never changed

2 sources

On Saturday, October 17, 2026, Meta removed language from its commercial terms that legal experts had read as permission to scrape publishers' sites for AI training. Per The Verge (Terrence O'Brien, 15:04 UTC October 17): 'After backlash, Meta is updating its terms again to remove language that suggested it could scrape content from sites, including news sites, to train its AI. Specifically, the following language no longer appears on the commercial terms page' — followed by The Verge's rendering of the removed language, opening with the outlet's own leading ellipsis: '… 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.' Per The Verge: 'Many, including Neiman Lab and James Grimmelmann, a professor of digital and information law at Cornell, interpreted it as an attempt to give Meta the right to use content scraped from sites to train its AI products for free and supersede blocking instructions in a site's robots.txt file.' Meta disputed that reading before removing the language: per The Verge, 'Andy Stone, Meta's vice president of communications, and spokesperson Daniel Roberts pushed back, saying that "There is no change to how Meta uses web crawlers; any suggestion to the contrary is false."' After the removal, per The Verge, Roberts said: 'We've heard you and want to clarify: the crawling terms we consolidated only applied to our ads and business products, and we've clarified that in the terms themselves. It does not change how Meta uses web crawlers. As always, content owners can use robots.txt to signal their preferences. This is not about scraping news sites, and these updates don't give us new rights that supersede those preferences.' Per TechCrunch (Anthony Ha, 21:31 UTC October 17): 'According to legal experts interviewed by Nieman Lab, this looked like a "remarkable provision" that could give Meta permission to scrape the website of any publisher with a Meta business account' — with the old language rendered using TechCrunch's own bracketed insertion '[a business account's]'.

Why now

The terms story every creator with a Meta business account should know about — because the terms language came down on Saturday (October 17) and the record around it is fully checkable. Per The Verge (October 17): 'After backlash, Meta is updating its terms again to remove language that suggested it could scrape content from sites, including news sites, to train its AI.' Second layer, the interpretation that started it: per The Verge, experts including Cornell professor James Grimmelmann read the language 'as an attempt to give Meta the right to use content scraped from sites to train its AI products for free and supersede blocking instructions in a site's robots.txt file' — and per TechCrunch, legal experts interviewed by Nieman Lab called it 'a "remarkable provision"'. Third layer, Meta's side is on the record and stays there: 'There is no change to how Meta uses web crawlers; any suggestion to the contrary is false.' — per Andy Stone and Daniel Roberts via The Verge — and Roberts' post-removal statement says the terms 'only applied to our ads and business products' and 'don't give us new rights that supersede those preferences.' Fourth layer, the actionable part is Meta's own line: 'As always, content owners can use robots.txt to signal their preferences.' — one check every site owner can run, whatever the legal fight settles.

Why it is worth publishing

The platform-policy card most directly about creators' own content: the story ran from a terms update to an outcry to a removal, and the record now contains the experts' reading, Meta's denial, and the removed text — three layers most coverage will collapse into one. The differentiation play is keeping the dispute a dispute: attribute the scraping reading to the experts, keep Meta's 'no change' framing attached, and never write that Meta's terms allowed scraping or that any site was scraped — no opened source supports either. Creators who publish the three-layer version own the platform-policy beat; everyone else will ship either 'Meta can scrape your site' or 'nothing ever happened', and the record supports neither.

Evidence basis

Two opened sources: The Verge (Terrence O'Brien; article:published_time meta 2026-10-17T15:04:36+00:00, Saturday calendar-verified; opened and read in full via raw HTML this run, 2026-10-19) and TechCrunch (Anthony Ha, In Brief; published_time meta 2026-10-17T21:31:32+00:00, Saturday; opened and read in full via raw HTML; published after The Verge and citing it). Calendar check: October 17, 2026 = Saturday; October 19 (this card's publish date) = Monday. Unopened origins, recorded per the 09-18 precedent: the Meta commercial terms page (facebook.com/legal/commercial_terms_preview, linked by The Verge; facebook.com unreachable from this run's network — connection failure, so the removal fact rides The Verge's opened report); the Nieman Lab article that first flagged the language (Cloudflare challenge this run; its facts ride the two opened outlets); Andy Stone's X post (unopened; the pushback quote rides The Verge). Per-outlet rendering discipline: TechCrunch renders the old language with its own bracket '[a business account's]' where The Verge's rendering reads 'such domains' behind the outlet's own leading ellipsis — never blended into a third form; TechCrunch leaves the spokesperson unnamed while The Verge names Daniel Roberts. The card never states that Meta's terms allowed scraping, never states that Meta scraped any site, and keeps Meta's dispute attached wherever the experts' reading appears. The web-search quota was exhausted this run — sources were reached by fetching listing pages and articles directly (09-08 fallback precedent).

“On Saturday, Meta removed the terms language that legal experts read as permission to scrape sites for AI training — while Meta insists its crawlers never changed.”

Angle

Frame it as 'the terms changed, the internet pushed back, and the language came out — here's the full three-layer record' in three beats. Beat one, what was there: per TechCrunch, Meta's updated commercial terms declared the company can 'use any content from [a business account's] domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of' its products — and per TechCrunch, legal experts interviewed by Nieman Lab called it 'a "remarkable provision"'. Beat two, the walk-back: per The Verge, the language 'no longer appears on the commercial terms page' as of Saturday (October 17), and Meta's statement says the crawling terms 'only applied to our ads and business products'. Beat three, what you can do: Meta's own line — 'As always, content owners can use robots.txt to signal their preferences.' — plus the honest boundary that the experts' reading was an interpretation Meta disputes, and no opened source says any site was scraped.

Format

Short talking-head video

Demo idea

A three-document walkthrough: show the experts' reading (Grimmelmann and the Nieman Lab interviewees, per The Verge and TechCrunch), then Meta's two statements (the 'no change' pushback and the post-removal clarification), then your own site's robots.txt with the one line that signals your preference — closing on what the record does not contain: any evidence a site was actually scraped.

Platform notes

The removal stays dated and past-tense — 'Meta removed the language on October 17', never present-tense 'Meta's terms allow scraping'. The scraping reading stays attributed to the experts (Grimmelmann, Nieman Lab's interviewees per The Verge and TechCrunch) with Meta's dispute attached ('There is no change to how Meta uses web crawlers; any suggestion to the contrary is false'); never write that Meta admitted wanting to scrape — Meta frames the removal as a clarification of consolidated terms. No claim that any site was scraped or that creators' content was used for training — no opened source supports it. robots.txt stays two-layer: Meta's own statement plus the experts' supersession reading, never an independent legal verdict. The two outlets' renderings of the removed language stay separate — The Verge's 'such domains' behind its leading ellipsis, TechCrunch's '[a business account's]' bracket — never blended. The removal receipt is The Verge's opened report; the live terms page was unreachable this run, so never say 'we checked the live page'. zh scripts keep Meta, robots.txt, Nieman Lab in English, write 移除 — never 道歉 or 认错 — keep 坚称/澄清 on every Meta sentence, and keep 法律专家解读 attributed.

Usable claims

On Saturday, October 17, 2026, Meta removed language from its commercial terms that legal experts had read as permission to scrape publishers' sites for AI training. Per The Verge (Terrence O'Brien, published 15:04 UTC October 17): 'After backlash, Meta is updating its terms again to remove language that suggested it could scrape content from sites, including news sites, to train its AI. Specifically, the following language no longer appears on the commercial terms page' — followed by The Verge's rendering of the removed language (which opens with the outlet's own leading ellipsis): '… 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.' Per The Verge: 'Many, including Neiman Lab and James Grimmelmann, a professor of digital and information law at Cornell, interpreted it as an attempt to give Meta the right to use content scraped from sites to train its AI products for free and supersede blocking instructions in a site's robots.txt file.' Meta disputed the scraping reading before the removal: per The Verge, 'Andy Stone, Meta's vice president of communications, and spokesperson Daniel Roberts pushed back, saying that "There is no change to how Meta uses web crawlers; any suggestion to the contrary is false."' After the removal, per The Verge, Roberts said in a follow-up statement: 'We've heard you and want to clarify: the crawling terms we consolidated only applied to our ads and business products, and we've clarified that in the terms themselves. It does not change how Meta uses web crawlers. As always, content owners can use robots.txt to signal their preferences. This is not about scraping news sites, and these updates don't give us new rights that supersede those preferences.' Per TechCrunch (Anthony Ha, 21:31 UTC October 17, an In Brief post citing The Verge): 'Meta recently updated its commercial terms of use, declaring that the company can "use any content from [a business account's] domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of" Meta products and services' (the bracketed insertion is TechCrunch's own), and 'According to legal experts interviewed by Nieman Lab, this looked like a "remarkable provision" that could give Meta permission to scrape the website of any publisher with a Meta business account and use the content to train the company's AI models.'

Evidence pipeline

From the news

Breakdown

A walked-back policy story where the discipline is keeping three layers separate. Layer one, the removal is dated and settled: per The Verge (October 17), 'the following language no longer appears on the commercial terms page' — the story is past tense, never present-tense 'Meta's terms allow scraping'. Layer two, the reading belongs to the experts: James Grimmelmann and the legal experts Nieman Lab interviewed (per The Verge and TechCrunch) interpreted the language 'as an attempt to give Meta the right to use content scraped from sites to train its AI products for free and supersede blocking instructions in a site's robots.txt file' — an interpretation, disputed by Meta in the same record ('There is no change to how Meta uses web crawlers; any suggestion to the contrary is false'), and Meta's post-removal framing ('only applied to our ads and business products') is a clarification claim, never an admission. Layer three, the record contains no scraped-site fact: interpretation, denial, removal — no evidence any site was scraped or any creator's content was used for training, so nothing of the kind is written. Layer four, the receipt is bounded: the live terms page was unreachable this run (facebook.com connection failure) and Nieman Lab's article was Cloudflare-blocked, so the removal fact rides The Verge's opened report — never 'we checked the live page'. Layer five, the renderings stay per-outlet: The Verge's 'such domains' behind its own leading ellipsis, TechCrunch's '[a business account's]' as the outlet's own bracketed insertion, TechCrunch's unnamed spokesperson beside The Verge's named Daniel Roberts — each quoted as its outlet's, never blended. Editor's rules: removal stays past tense, interpretation stays attributed, the dispute never leaves the sentence, and the receipt stays the outlet that opened the page.

Sources

Risks

Before publishing, re-check each layer: the removal stays dated and past-tense; the experts' reading stays attributed and Meta's denial stays attached; no scraped-content claim anywhere; robots.txt stays two-layer attributed; the per-outlet renderings stay unblended; and the receipt stays The Verge's opened report, never a claim about the live page. If your script compresses any of these, cut the detail rather than round it.

Demo ideas

Three-document walkthrough: the experts' reading, Meta's two statements, and your own robots.txt — closing on what the record does not contain
Timeline card: terms language flagged by Nieman Lab (per TechCrunch and The Verge), Meta's 'no change' pushback, The Verge's removal report Saturday October 17 — three beats, each with its own date and attribution