Meta has rewritten its commercial terms once again, this time removing language that appeared to grant the company permission to scrape content from outside websites — including news publishers — and use it to train its artificial intelligence systems. The rollback followed sharp backlash from publishers and legal observers. For a company that has signed paid AI licensing deals with media outlets, the original wording looked a lot like a way to skip the paywall entirely.

What changed in Meta’s terms

The deleted passage was broad in scope. It covered content from domains, web pages, or URLs connected to Meta’s commercial services, and it listed a long set of permitted uses.

“Meta may use any content from such domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of, the Meta Company Products, services, and features provided under these Commercial Terms and any other applicable supplemental terms.”

That clause no longer appears on the commercial terms page. The word doing the heavy lifting was “train” — precisely the term that drew attention from publishers and legal experts.

The revised Commercial Terms carry an effective date of October 30, 2026. As first written, they authorized a wide range of automated tools to crawl and scrape data from URLs tied to business accounts across Meta’s platforms, which include Facebook, Instagram, WhatsApp, and Messenger. These Commercial Terms set the ground rules for the businesses and advertisers operating across Meta’s apps, so shifts in their wording affect more than Meta alone.

Meta responded with a clarification around October 16–17, 2026. The updated version limits the scraping permissions to advertising and business products. It also strips out the language allowing AI training. In addition, Meta affirmed that it respects existing web protocols such as robots.txt — the long-standing web standard that lets site owners signal which automated crawlers may access their pages.

Why publishers reacted so strongly

The News/Media Alliance, an industry group representing publishers, called the initial terms “outrageous” and described them as a potential violation of legal standards. Other industry voices went further, calling the original terms an abuse of market power. That argument rests on how dependent many newsrooms are on Meta’s platforms for distribution.

The licensing angle sharpened the complaint. Meta has established paid AI licensing agreements with media outlets including CNN and Fox News — deals that show the company is willing to pay for news content in some cases. Legal experts suggested the original wording could enable scraping news sites for AI training potentially worth “many millions” in licensing fees. That made the original terms look like a potential free alternative to the very contracts Meta has been signing.

What this means for Meta, publishers, and the AI market

For publishers, the revision is a meaningful but narrow win. The AI training language is gone, scraping permissions confined to advertising and business products, and robots.txt is acknowledged. Since Meta has rewritten these terms more than once, the definitive wording that takes effect on October 30, 2026 is the key detail to watch from here.

For Meta, the episode carries reputational cost. The company relies on news outlets as both platform users and, in some cases, paid licensing partners.

The speed of Meta’s reversal stands out. The clarification arrived before the October 30 effective date, suggesting the company preferred to fix the wording rather than defend it through a prolonged public fight.