“Does Meta think we can’t read?” Meta abandons sketchy commercial terms update.

After weeks of backlash from publishers, courts, and regulators, Meta told Reuters it is pausing the commercial-terms changes and keeping existing terms in effect while it "engage[s] with stakeholders." The rewrite would have required Meta Business account holders — including nearly every news publisher with a Facebook page — to allow extensive AI training on external sites linked from their pages.

Legal experts warned Nieman Lab the vague language could let Meta ignore robots.txt preferences and crawl off-platform content at scale, potentially sidestepping licensing fees. The terms even required publishers to create test credentials so Meta bots could bypass paywalls. There was no opt-out short of abandoning a business Facebook page.

This is not a quiet product tweak. It is a fight over whether platform power can be rewritten into a training-data pipeline.

We're calling this what it looked like from day one: a terms-of-service grab dressed as a partnership update. Businesses that need a Facebook page were told to accept crawling of anything they link — or leave. That is leverage, not consent.

The pushback was unusually broad. Courts in Brazil and Norway issued preliminary injunctions. The UK Competition and Markets Authority opened a formal investigation into whether Meta was exploiting a dominant position to seize AI training data. Sens. Adam Schiff and Marsha Blackburn both asked Mark Zuckerberg to withdraw the terms. Theo Bamber of the UK News Media Association asked the question that stuck: "Does Meta think we can't read?" Softening "train" into "develop" fooled nobody when the scrape still fed "any business product."

Meta's Andy Stone said the company would consult industry groups and delay past the November 20 date if needed. A spokesperson now says the pause is indefinite. Meta still insists crawling practices are unchanged and that robots.txt preferences remain available — yet Nieman Lab notes the revised terms still lacked specific language committing to those preferences.

I want AI companies to license what they need and respect what they cannot take. Useful models do not require hostage terms aimed at every publisher's homepage. The downside if this pause is just time-buying: the same architecture of "accept or lose the page" returns with cleaner verbs and the same power imbalance. Watch the consultation for hard robots.txt commitments, real opt-outs, and whether regulators treat this as a one-off climbdown or a pattern.

The changes were first set for October 30, then pushed to November 20 amid global scrutiny. Meta had already faced drawn-out legal fights with publishers over unwanted AI scraping before this commercial-terms fight landed.