Meta removes language that may have allowed it to scrape news sites
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:
Meta has quietly edited its commercial terms, deleting a clause that appeared to grant the company broad rights to scrape content from news sites and other domains for AI training. The removal follows public backlash and pointed commentary from observers like the Neiman Lab and Cornell law professor James Grimmelmann, who flagged the language as a potential green light for unchecked data harvesting.
On its face, this is a retreat. The deleted sentence explicitly allowed Meta to use content from any domain to train, develop, and improve its products. Its absence suggests the company heard the criticism and decided the reputational cost outweighed any legal advantage. But the move is less a concession than a recalibration. Meta has not renounced scraping; it has simply removed the most visible admission of intent.
The practical effect is minimal. Meta's ability to scrape public web data does not hinge on a single terms page. The company can still rely on broader legal arguments, such as fair use or the public nature of the content, depending on jurisdiction. What changes is optics. By deleting the clause, Meta avoids handing regulators and publishers a written confession that could be cited in litigation or policy debates.
This is a familiar pattern in the AI arms race. Companies push the boundary of what their terms permit, gauge the reaction, and then retreat just enough to preserve deniability. The underlying strategy remains unchanged: train on as much data as possible, and only concede ground when the noise becomes too loud.
For publishers, the signal is mixed. The removal offers temporary relief, but it does not establish a durable barrier. The legal landscape around AI training data is still being written, and Meta's edit is a tactical move, not a policy shift. The real test will come when a court or regulator forces a definitive answer on whether scraping for AI training is permissible without explicit consent.
Until then, this is a story about language, not behavior. Meta has cleaned up its terms, but the machinery underneath is still running.