By Hamza Faisal
Meta has quietly edited its updated commercial terms to delete the sentence that legal experts read as a free pass to scrape news websites and train its AI models on their content — ending, for now, one of the ugliest fights between Big Tech and the press this year.
The removed language had said Meta could use "any content" from domains, web pages, or URLs connected to its commercial services to "provide, develop, improve, research, train, and maintain" Meta Company Products. The word doing the heavy lifting was "train" — and publishers heard it loud and clear.
The clause sat inside commercial terms set to take effect on October 30, covering business use of Facebook, Instagram, WhatsApp and Messenger. Almost every major news outlet keeps a Meta business account to manage pages, promote stories and run ads — which, under the experts' reading, could have put their entire websites within Meta's reach. There was no targeted opt-out: reject the terms and risk losing commercial access to the platforms.
Backlash forced the edit
Nieman Lab first flagged the terms, and Cornell digital-law professor James Grimmelmann warned the contract said nothing about the robots.txt files sites use to block crawlers — and could be read as overriding them. The News/Media Alliance called the original provisions "outrageous" and a potential breach of legal standards.
Meta's first response was firm. Communications VP Andy Stone and spokesperson Daniel Roberts denied any change to how the company crawls the web, describing the update as a consolidation of existing commercial and ad terms. Then Meta went further than a clarification — it stripped the contested sentence from the page entirely.
"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," Roberts said.
Meta says the scope was always narrow
The company's position now is that the consolidated crawling permissions exist only to support its advertising and business products — a much narrower purpose than critics saw in the earlier text. Meta also reiterated that it respects robots.txt, the standard file sites use to tell crawlers what they may copy.
That leaves the central question untouched: whether training AI models on scraped web content requires permission at all. Meta already pays outlets including CNN and Fox News for AI training licenses, and courts and regulators are still writing the rules everyone else will have to follow.
A temporary win, not a settlement
For publishers, the edit is relief, not resolution. Deleting the clause avoids handing regulators and litigants a written statement of intent they could cite in court — but it does not establish that scraping for AI training is off-limits. Sites still have to maintain their robots.txt files, and Meta still crawls the web at planetary scale.
The pattern is becoming familiar: push the boundary in the fine print, gauge the reaction, and retreat just enough to preserve deniability. October 30 is still coming — only now with one fewer sentence to argue about.
By Hamza Faisal
