Regulation
Meta delays its new terms after news outlets complain about AI
Meta was going to apply new terms for businesses on October 30 that, according to several experts, opened the door to training its AI on news websites. It has pushed them back to November 20 and warns it could delay them further.

The essentials
- The text of the new terms said Meta could use content from the websites of business accounts to, among other things, "train" its products (The Verge, TechCrunch).
- After the criticism, Meta removed that sentence and the word "train" from its terms (The Verge, Ars Technica).
- Claimed by one partyAccording to Cornell professor James Grimmelmann, the mention of websites "associated" with the business would still allow Meta to copy entire sites (The Verge).
- Claimed by one partyMeta has delayed the date the terms take effect from October 30 to November 20 (The Verge).
- Claimed by one partyAndy Stone, Meta's head of communications, says the company does not plan to make changes without first consulting the industry in depth (The Verge).
Why it matters
What is at stake is money. According to Ars Technica, news outlets could charge Meta millions for letting it use their articles to train its AI. News outlets are already fighting OpenAI in court over whether that kind of unlicensed use is legal. The first version of these terms could have given Meta broad permission, whatever the outcome of those cases (Ars Technica).
The delay suggests that pressure from news outlets works, but the text is not final yet. If the experts are right, removing one word changes little in substance. The outcome of the consultation Meta says it plans will show whether the change is real.
If you have a business page on Facebook or Instagram linked to your website, these terms affect you. According to Ars Technica, they don't let you opt out unless you stop using Meta (Ars Technica).
The details
According to Ars Technica, the new terms apply to all business accounts and don't allow anyone to opt out. The only way out suggested is to stop using Meta's services (Ars Technica). The experts Nieman Lab spoke to also feared that Meta would argue its contract outweighs robots.txt. That is the file a website uses to tell automated programs which parts they may not read (The Verge).
Meta has responded several times. First, its spokesperson Daniel Roberts said the company was only bringing its advertising terms together (The Verge). Then, in a statement reported by TechCrunch, Meta said those rules only applied to its ads and business products. It also said they don't change the way its programs crawl the web (TechCrunch). A spokesperson confirmed to Ars Technica that Meta did not notify users when it edited the text (Ars Technica).
Removing the word "train" has not been enough to reassure the experts. James Grimmelmann, a professor of digital law at Cornell, told Nieman Lab that the sentence about websites "associated" with the business's products or services remains untouched. In his view, that sentence would let Meta copy entire sites (The Verge).
Meta ended up pushing back the date the terms take effect, originally set for October 30, to November 20. According to Roberts, it wants more time to explain to its partners what is changing and what isn't (The Verge). Later, Meta's head of communications, Andy Stone, wrote on X that the company does not plan to make changes without first consulting the industry in depth. Meta admits it could move the date beyond November 20, according to The Verge (The Verge).
What we don't know
- What the final date for the terms to take effect will be and what the final text will look like.
- Whether Meta will use the mention of "associated" websites to copy content to train its AI.
- Whether Meta's crawling programs will respect the blocks that websites put in place.
Background
This is not the only dispute Meta faces over the material it uses to train its AI. According to TechCrunch, Meta won a high-profile copyright infringement lawsuit last year, but other cases against AI companies, including Meta, are still open (TechCrunch). We have already reported that a judge is requiring it to identify which executives approved downloading pirated books to train its AI.
How this story has changed
- Meta has delayed the date the terms take effect to November 20 and warns it could push it back further until it has consulted the industry.
- Meta has removed the word "train" without notifying users, and experts argue the change is not enough because another sentence still allows it to copy entire websites.
- After the criticism, Meta has removed from its terms the paragraph about using its customers' websites and says it only applied to its ads and business products.
- We published the story.
Sources
- Meta’s new terms may allow it to train AI on news sites for free
- Meta may use sweeping new commercial TOS to avoid millions in AI licensing fees
- Meta removes language that may have allowed it to scrape news sites
- Meta removes language that raised scraping and AI training concerns
- Meta drops the word "train" from its commercial TOS after AI backlash
- Meta’s AI could still scrape news sites, legal experts say
- Meta delays new business terms after publisher pushback on AI scraping
- Meta says it’ll probably delay its controversial new business terms again
Was this story useful? Yes Not really