Meta's New Terms May Let It Train AI on News Sites Without Asking
A clause taking effect October 30th appears to give Meta permission to scrape websites linked to businesses that use its platforms, including publishers who never agreed to that deal.
Key points
- Meta's updated terms of service, effective October 30, 2026, include language that may allow it to crawl and scrape external websites "associated" with businesses that use its platforms.
- The new terms state Meta can use that outside content to "develop, improve, research" and "train" its AI systems.
- Publishers and online news outlets that maintain a Facebook or Instagram presence could fall within scope, even if they never agreed to provide training data.
- Publishers have very little time to challenge the terms, remove their business pages, or accept the new arrangement by default.
Meta is about to change what it can legally do with content it finds on the web, and the people most affected may not realise they're in the picture at all.
The company's revised terms of service, first flagged by Nieman Lab and reported by The Verge AI, take effect on October 30, 2026. A key clause says Meta can use content from websites "associated" with businesses on its platforms to develop and train its AI. If you run a news site and maintain a Facebook page, Meta may now claim the right to harvest your articles and feed them to its models.
What does "associated" actually mean?
That word is doing enormous work, and nobody outside Meta's legal team knows exactly how far it stretches. A small bakery with an Instagram account probably has a website. So does every newspaper, hospital, and retailer with a business profile on any Meta platform.
The terms don't appear to cap what kind of business, what size of site, or what category of content qualifies. That breadth is what has alarmed publishers: the clause looks less like a narrow data-licensing arrangement and more like a blanket pass for industrial-scale web scraping, funded indirectly by the businesses Meta already charges for advertising.
Why does this matter for readers?
For most people reading the news online, the immediate effect is invisible. The consequences compound, though. If AI systems train on journalism without payment or permission, the financial case for producing that journalism weakens. Fewer reporters means thinner coverage of hospitals and local government, including the health stories this site covers closely.
This isn't an isolated move. When we covered Australia's proposed opt-out AI training rules on 17 September, the same structural problem surfaced: consent defaults to "yes" unless someone actively objects. Meta's new terms follow the same logic, applied by a private company rather than a government.
Who consented, and who benefits? Those questions have no good answers here. Meta needs vast amounts of fresh data to keep its models competitive, and it's now writing its own rules about where to find it.
October 30th is close. Silence, under these terms, looks a lot like consent.
