Meta has revised its business terms after publishers objected to language that appeared to allow the company to train AI models on content from websites linked to business accounts. The new wording removes an explicit reference to training, but legal experts say it may still leave room for Meta to use scraped material in developing its products.
What changed in Meta’s terms
The business terms are set to go into effect October 30th. Their earlier version said Meta “may use any content” from websites associated with a business account to “develop, research” and “train” its products.
That language raised concerns among news organizations and legal experts. They said it could give Meta an argument that its commercial terms take precedence over other instructions that restrict scraping, including a website’s robots.txt file.
After fierce backlash from the publishing industry, Meta removed the reference to training. The revised terms still refer to using scraped content to develop a “business product,” however, and experts say the scope of that phrase is unclear.
Why experts say the concern remains
James Grimmelmann, a professor of digital and information law at Cornell Law School, told Nieman Lab that the revision does not address language about content “associated with your products or services.” In his view, that wording could still let Meta scrape entire websites.
Matt Topic, a media and intellectual property attorney at Loevy & Loevy, also told Nieman Lab that the new terms may not close the door to AI training. He said “business product” does not appear to be defined, and argued that developing a product could be broad enough to include training it.
The distinction matters because the revised wording no longer names training directly, while leaving a broader permission to develop products. The experts’ concern is that this broader language may cover some of the same activity even without using the word “train.”
Meta says website preferences still apply
Meta has said the updated terms are not intended to let it scrape news content for free. In a statement to The Verge, spokesperson Daniel Roberts said content owners can use robots.txt to signal their preferences.
Roberts also said the update “is not about scraping news sites” and does not give Meta new rights that override those preferences. That position differs from the experts’ concern about how the revised business terms could be interpreted.
The dispute therefore centers on what the terms permit and how they interact with website instructions. Meta says those instructions remain effective; the legal experts cited by Nieman Lab argue that the broad language about developing products may still leave uncertainty.
The unresolved question for publishers
For publishers, removing the explicit training language is a change, but it has not ended the debate. The remaining wording is the focus: whether permission to use content associated with a business account to develop a product could extend to AI systems trained on material gathered from those sites.
The source does not establish that Meta is currently using scraped news content to train AI under the revised terms. Instead, it describes competing interpretations: Meta says it has not gained new rights over website owners’ preferences, while legal experts say the revised language may still be broad enough to support scraping and product development.
Until the scope of “business product” and the relationship to robots.txt are clearer, the terms leave publishers with a question about whether their existing instructions will be respected in practice.