Meta's revised terms still leave room for AI training on news content, legal experts warn
Despite removing explicit 'train' language, Meta's updated business terms contain ambiguous language that could permit scraping news sites for AI development.
Last verified:
Semantic Loopholes in Meta’s Revised Terms
Meta’s attempt to assuage publisher concerns by removing the word “train” from its October 30 business terms update may have failed to address the underlying legal ambiguity. According to The Verge, the revised language still permits broad use of scraped content through strategically undefined terminology.
The original terms stated that Meta “may use any content” from business account websites to “develop, research” and “train” Meta’s products. Following fierce industry backlash, Meta excised the explicit reference to training. However, James Grimmelmann, a professor of digital and information law at Cornell Law School, tells Nieman Lab that this revision is insufficient: “This does nothing to restrict the ‘associated with your products or services’ [language] that would let Meta scrape entire sites.” The phrase remains in the updated terms, according to Grimmelmann’s analysis.
The Problem with “Develop” and “Business Product”
Matt Topic, a media and intellectual property attorney at Loevy & Loevy, argues that the revised terms’ use of “develop” and “business product” may be capacious enough to encompass AI training despite the removal of the explicit word. According to The Verge, Topic told Nieman Lab: “The revised version of the new terms still allows Meta to use that scraped content to develop any ‘business product.’ That doesn’t seem to be a defined term, and presumably Meta is developing AI business products. I think ‘develop’ is broad enough to include ‘train.’”
The substitution of “develop” for “train” suggests Meta’s revision was terminological rather than substantive. If a court were to interpret the terms, the absence of a legal definition for “business product” could permit Meta to argue that AI model development qualifies as a permissible use.
Meta’s Position and the robots.txt Question
In a statement to The Verge, Meta spokesperson Daniel Roberts stated that the company respects robots.txt signals, a standard mechanism publishers use to instruct scrapers about crawl access. “Content owners can use robots.txt to signal their preferences,” Roberts said. He added that the updated terms “is not about scraping news sites, and these updates don’t give us new rights that supersede those preferences.”
However, the legal concern centers on whether Meta’s business terms could be interpreted as superseding robots.txt altogether. According to The Verge’s reporting, legal experts initially flagged the possibility that Meta might argue its commercial terms override historical scraping conventions.
Why This Matters
The ambiguity in Meta’s revised language creates a practical enforcement problem for publishers. If publishers cannot rely on robots.txt or unambiguous contractual language to prevent scraping for AI training, they lose one of their primary defenses against unauthorized use of their content. This may force publishers to pursue costly litigation to clarify their rights—or accept that Meta can extract value from their journalism without compensation. For news organizations already struggling with revenue, this uncertainty underscores the need for clearer legal frameworks around AI training data sourcing.
Frequently Asked Questions
What did Meta change in its business terms?
Meta removed explicit language stating it would 'train' its products on content from business account websites, responding to publishing industry pushback. The revised terms take effect October 30.
Can Meta still scrape news sites under the new terms?
According to legal experts quoted by The Verge, yes. The terms still allow Meta to 'develop' 'business products' using scraped content, and 'develop' may be broad enough to encompass training AI models.
What is robots.txt and does it protect publishers?
Robots.txt is a standard file that instructs web scrapers which pages they can access. Meta claims it respects robots.txt preferences, but legal experts worry the new terms could supersede those signals.
What do the legal experts disagree with?
Experts argue that undefined terms like 'develop' and 'business product' in Meta's revised language are vague enough to permit AI training, even though the word 'train' was removed.