Meta TOS AI Training Update Faces Legal Concerns

The updated commercial terms would require business accounts to let Meta train AI on content from their linked websites, not just scrape their social media posts.

The update offers no opt-out. Businesses that refuse the terms are simply told to stop using Meta services. For news publishers, the stakes are enormous. They could charge Meta millions in licensing fees to continually access their articles for AI training. News organizations are already fighting OpenAI over whether training without a license counts as fair use. The Meta TOS AI Training terms appear designed to sidestep that legal battle no matter which way courts lean.

Sweeping Permissions With No Opt-Out

The core of the update reads like a grant of near-total access. Meta could send automated software onto any domain, page, or URL tied to a business account:

You authorize us (or any entity acting on our behalf) to use automated software (such as web scrapers, bots, robots, spiders, crawlers, and user agents) and create test credentials to access, index, cache, analyze, collect, retrieve, or crawl the domains, web pages, or URLs (and any content available through them) that are associated with your products or services.

The word "AI" never appears. But the single use of "train" combined with the exhaustive list of scraping tools makes the intent plain. Meta may use any content gathered this way to "provide, develop, improve, research, train, and maintain" its products and services.

The Scraping Toolbox

The terms explicitly name every tool Meta would deploy:

• Web scrapers
• Bots and robots
• Spiders and crawlers
• User agents and test credentials

Why "Test Credentials" Worries Experts

Three legal experts said it is unclear whether a court would enforce the terms. But the language could be used to argue that Meta can scrape content from news sites and feed it into AI products. The phrase "test credentials" troubles them most, because it could allow Meta to bypass paywalls and automate scraping.

Meta pushes back. A spokesperson said there is no change in how the company uses web crawlers and called any suggestion to the contrary false. Content owners can still manage access through robots.txt. The company also said it has no plans to use the terms as a legal basis to ignore those instructions.

News Publishers Stand to Lose Millions

James Grimmelmann, a professor of digital and information law at Cornell Law School, said Meta could "train AI on any content from their sites" and possibly "deploy some of that content, even in competition with the publishers themselves" on Meta platforms.

Matt Topic, a media and intellectual property attorney at Loevy & Loevy, called it "a remarkable provision to bury in terms of service," given that this would hand Meta access to material that would likely cost many millions of dollars in a licensing market.

The Meta TOS AI Training update lands at an awkward moment for publishers who are already weighing whether to stay on Meta platforms at all.

Meta Calls It Streamlining. A Comparison Says Otherwise.

Meta announced the update in a blog where the company claimed it was simply streamlining its commercial terms. A spokesperson said language around crawling had been scattered across different terms of use, including terms about generative AI in ads, and the update consolidates them in one place.

A direct comparison of the current terms and the proposed updates shows fresh language and broader provisions. The term "test credentials" does not appear in any of the policies Meta says it is consolidating:

• Meta's current commercial terms
• Product Catalog Terms
• Ad Creative Generative AI Terms

"Test credentials" appeared only this month in Self-Service Ads Terms. When asked about prior policies, a Meta spokesperson pointed to a Seller Agreement that was not mentioned in the blog announcing the update. That agreement requires e-commerce sellers to let Meta retrieve information from their URLs so the company can improve listings. Legal experts likened the approach to the antitrust concept of "tying," where a business forces a buyer to accept a second product as a condition of buying the first. They compared it to Google's prior practice of requiring websites to opt in to scraping for AI Overviews if they wanted to be indexed for general search. Google has since started allowing opt-outs.

One Publisher Is Already Weighing an Exit

October 30 could trigger a broader scale-back among news organizations. That's the fear. Stephen Beale, founder and editor of The Steampunk Explorer, depends on Facebook for traffic, and while he knows his website has already been "scraped to death," the update felt like "such an arrogant, in-your-face move" on Meta's part.

The frustration is compounded by timing. Meta has just begun charging users to post links on its platforms, meaning publishers could first pay to share links and then see Meta claim permission to scrape those same links for training. The Meta TOS AI Training terms could hand the company a path to charge publishers and then harvest their content in one motion.

Should courts affirm that training on non-news articles is fair use, publishers question whether Meta's stance on robots.txt might shift. The terms, written now, would already be in place to grab content without a new legal fight.