Analysis
Meta is rewriting the rules for the businesses that depend on it. A revised set of commercial terms of service, effective October 30, authorizes Meta "or entities acting for it" to run crawlers and bots against the pages, apps and ad accounts of any business customer -- and to use what they collect to develop, improve and train its products, including AI models, according to The Register. Meta defines a "business customer" broadly: anyone running ads, selling products, managing a Page or Group, or using tools like Ads Manager or Meta Business Suite. By Meta's own account, more than a million businesses already use its AI agents inside WhatsApp and Messenger alone, and the new terms sweep in Instagram, Facebook and its advertising stack as well.
This isn't Meta's first run at claiming training data by fiat. Pulse covered Meta's push to ship ever-larger open models even as the company works through a history of data controversies that include the Cambridge Analytica scandal and a social-media-addiction settlement with 52 state attorneys general -- the backdrop Santa Clara University law professor Eric Goldman cited when he said Meta "has no reservoir of good faith" with the people it's asking to accept the new terms.
What The Terms Actually Change
The practical shift is leverage: continued access to ads, messaging tools and a Page's own audience is now bundled with permission to scrape that Page's content, with no visible way to decline while keeping the account. Danielle Coffey of the News/Media Alliance called the language "outrageous and probably illegal," arguing that conditioning basic platform access on surrendering intellectual property is an abuse of market power rather than a negotiated exchange. Goldman's concern is narrower but just as practical: reports suggest Meta will let businesses signal scraping preferences through robots.txt, but the terms themselves don't say so, and he notes the law doesn't require anyone to honor robots.txt in the first place.
Meta isn't alone in treating training data as something to claim rather than license. OpenAI and Anthropic are both fighting publisher lawsuits over how their models were trained, and Google and X have each rewritten terms of service in the past two years to widen what they can pull from their own platforms. What's different here is scale and structure: Meta is the first to fold scraping rights directly into the commercial terms that a million-plus businesses must accept just to keep running ads or using WhatsApp for customer service.
Critics argue the real test isn't the terms themselves but whether anyone forces Meta to honor a stated preference. Meta hasn't said what happens to a business that blocks its crawlers outright, and it didn't respond to The Register's request for comment. Nieman Lab's reporting that nearly every news publisher holds at least one Meta business account suggests the terms could sweep in exactly the content owners most likely to object -- and most able to afford a lawsuit. However, until a publisher or a regulator tests the "probably illegal" framing in court, the terms stand as written, effective October 30.
