Meta removes language that raised scraping and AI training concerns

Meta recently consolidated its commercial terms, and one line said it could use content from a business account's domains, web pages or URLs to develop, research and train its products. Legal experts quoted by Nieman Lab flagged it as a provision that could let Meta train AI on the site of any publisher with a Meta business account. A Cornell law professor was among those who read it as potentially overriding a site's robots.txt preferences.

After the backlash, Meta removed the language. Its spokespeople insist nothing ever changed about how Meta crawls the web. The company now says the consolidated crawling terms applied only to its ads and business products, and that robots.txt remains how content owners signal their preferences.

For publishers, that's a real concession on paper. For anyone who signs platform terms to run ads, it's a reminder to read them.

We'll take the fix. Meta heard the objection, removed the clause, and put robots.txt back at the center of its public statement. That's the right outcome, and it came fast.

Still, we don't think the backlash was overblown. Terms of service are where rights actually live. A promise from a communications team is nice, but a sentence in a contract is what gets cited later. When a company that is training large models adds the word train next to the words any content from your domains, publishers are right to treat that as a claim, whatever the stated intent.

Our read is that this looks more like careless drafting during a consolidation than a secret land grab. Meta says the language only covered ads and business products. Maybe so. But the episode shows the real tension. Every publisher that wants Meta's ad tools has a business account, and if access to distribution quietly comes bundled with access to content, the business account becomes the consent form.

We want AI models trained on good journalism. We also want that to happen through licensing and clear opt-outs, not through fine print attached to an ad account. If this kind of language becomes normal across platforms, robots.txt stops meaning much, and smaller publishers who can't afford lawyers to read every update lose by default.

I'm watching for the next terms update, from Meta and from everyone else. The backlash worked this time because people noticed.

Meta won a high-profile copyright lawsuit over AI training last year, but other cases against AI companies, including Meta, are still moving. Robots.txt is a voluntary signal sites use to tell crawlers what they may access.