Meta’s AI could still scrape news sites, legal experts say

Meta's updated business terms first said it may use any content from websites tied to a business account to develop, research and train its products. After publishers pushed back, Meta deleted the training language. Legal experts told Nieman Lab the edit doesn't settle anything.

Cornell Law's James Grimmelmann pointed to language about content "associated with" a business's products as broad enough to cover scraping entire sites. Media lawyer Matt Topic of Loevy & Loevy argued that permission to develop an undefined "business product" is wide enough to include training. Meta's response is that robots.txt still governs and that the update gives it no new rights over those preferences.

This is a story about how much meaning one deleted word can carry, and the honest answer is not much. Contracts get read for what they permit, not for what the drafter says it meant. If "develop" can reasonably cover training, and "business product" has no definition, then removing "train" mostly changes the optics.

We'll give Meta the part of its argument that deserves credit. It says this isn't about scraping news sites and that robots.txt preferences still apply. That's the right principle. Machine-readable opt-outs that crawlers respect are one of the few things keeping the open web working as AI companies scale up. If Meta truly means it, the cleanest path is to write that commitment into the terms themselves and define the vague phrases lawyers are flagging.

The concern the experts raised earlier is the one that sticks with us: that commercial terms could be used to argue they supersede blocking instructions. That would turn a voluntary norm into something a platform can contract around, and every publisher that uses Meta's business tools would be on the hook without ever deciding to license anything. That's a quiet shift with large consequences.

Where this leads if it becomes normal is predictable. Every major platform rewrites its terms to claim a little more, publishers respond with harder blocks and paywalls, and the AI tools everyone uses get trained on a narrower, less reliable web. Good AI needs good sources. Treating those sources as a free input through fine print undermines the supply it depends on. I'm watching for whether Meta defines its terms or keeps leaning on statements to reporters.

Robots.txt is the file sites have long used to tell crawlers, including search engines like Google, what they may access. It is a convention, not a law. The worry with Meta's terms is less about the crawler itself and more about whether a separate contract could be cited as permission when a site's robots.txt says no. Meta has said its terms don't grant rights that override those preferences.