Meta drops 'train' from terms after backlash, but critics say AI training is not banned

Meta has rewritten its website crawling terms to remove language that would have allowed the company to train its artificial intelligence on the content of businesses using its advertising tools. The company made these changes late on Friday following complaints from business customers and media outlets, according to reports by Gizmodo.

The revised terms no longer use the word train when describing how Meta uses data gathered from websites linked to its business products. Meta stated in a message to the Mercury News that it heard customer feedback and wanted to clarify its intentions. The company emphasized that these crawling terms apply only to its ads and business products, a limitation it has now explicitly spelled out in the updated text.

Context on the Changes

The original clause permitted Meta to use automated software to access, copy, and re-crawl websites associated with its business products, according to PPC Land. That previous wording also covered using that scraped content to train Meta’s own products. The updated terms remove training from the list of permitted uses for the gathered data. However, the News/Media Alliance, which represents US publishers, argues that the fix is insufficient because it merely limits stated uses rather than banning them.

Sam Quigley, a spokesperson for the News/Media Alliance, told the Mercury News that training was taken out of the listed uses but not expressly banned. He noted that the terms still leave the possibility of training open. Quigley also pointed out that the terms do not state that Meta will respect robots.txt, which is the file sites use to tell crawlers which pages they may visit.

Who Is Affected

The terms cover data gathered by web scrapers, bots, spiders, crawlers, and user agents. This broad wording becomes significant as AI agents spread across the web and old rules for bots come under strain. Meta is pushing hard on AI after launching Muse, its chatbot, while startups such as Firecrawl are building web tools for agents. The terms take effect on 30 October. Businesses accept them simply by carrying on using Meta’s commercial products.

What should you do?

Businesses that use Meta’s advertising and commercial tools should review the updated terms before 30 October to understand how their website data may be accessed. Organizations concerned about AI training should verify if their current robots.txt files are being respected, as the new terms do not guarantee compliance. Companies should monitor for further clarifications from Meta regarding the specific uses of crawled data, given that training is not expressly banned.

What to do and how to stay safe: Meta

• Review the updated Meta business terms to understand how website data may be accessed and used.
• Audit your robots.txt configuration to ensure it clearly defines which pages crawlers may visit.
• Monitor official Meta communications for further clarifications on AI training and data usage policies.
• Assess your organization's risk tolerance regarding the potential use of public web content for AI models.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Did Meta completely ban AI training on crawled business websites?

No, Meta removed the word train from the terms but did not expressly ban training, according to the News/Media Alliance.

When do the new terms take effect?

The revised terms take effect on 30 October for businesses that continue using Meta’s commercial products.

Do the new terms guarantee Meta will respect robots.txt?

No, the terms do not state that Meta will respect robots.txt, which sites use to control crawler access.