Meta is backing off its new business terms again. These are the terms that seemed to let the company train its AI models on content from news sites that use Facebook and Instagram. According to The Verge AI, Meta now says rollout will probably slip past November 20th. That’s the date it already picked once, after pushing back the original October 30th launch.
The Short Version
- What happened: Meta says it will likely delay its updated ads and business terms beyond November 20th.
- Who said it: Andy Stone, Meta’s head of communications, in a post on X.
- Why: Publishers pushed back hard, and senators from both parties criticized the terms.
- What’s in effect now: Meta’s existing terms. Nothing changes until the consultation ends.
- The open question: Whether the final terms will clearly rule out AI training on publisher content.
What Meta Actually Said
Stone said Meta does “not plan to make policy changes without a full consultation with industry groups and leaders.” He added: “We’ve heard publishers’ feedback on our proposed ads and business terms update. We’re committed to getting this right, and as we expect this consultation could extend beyond November 20, in which case we’d delay implementation of the updated terms. Until then our existing terms remain in place.”
Notice what’s missing. There’s no new date. The Verge AI reports enforcement may now be delayed indefinitely. That’s a real change from a company that planned to flip the switch tomorrow.
How This Got Messy
The trouble started with wording. The first draft appeared to give Meta permission to “train” its products on content from sites in its business ecosystem. Publishers saw that and assumed the worst, which is fair.
Meta then removed the explicit training language. That didn’t settle it. Legal experts told Nieman Lab the revised version still doesn’t rule out AI training. So Meta pushed the start date to November 20th. Now that date is slipping too.
The pressure is coming from several directions:
- The News/Media Alliance, which represents a large share of US publishers, has been pushing back.
- Sen. Marsha Blackburn (R-TN) said the terms show Meta is “attempting to exploit American creators and businesses.”
- Sen. Adam Schiff (D-CA) urged Meta to reverse the change.
It’s rare for criticism to come from both parties on a tech policy issue like this. That probably explains why Meta isn’t just quietly running out the clock.
Why This Matters for AI
I think this story is about more than a terms-of-service fight. It’s about who pays for the data that trains AI models.
For the past few years, AI companies have taken two paths with publishers. Some sign licensing deals and pay for access. Others scrape first and end up in court. Meta’s approach looked like a third path: put the permission in the fine print of a business relationship publishers already depend on. If you want Facebook and Instagram traffic or ad tools, you agree to the terms.
That’s why publishers reacted so strongly. A clause like that could quietly turn content worth licensing into free training data. And because so many outlets rely on Meta’s platforms for distribution, they couldn’t easily walk away.
By delaying, Meta is basically admitting the move cost more goodwill than it was worth, at least for now.
What to Watch Next
If you run a publication, build AI products, or handle platform policy, keep an eye on these:
- The final wording. The key question is whether Meta explicitly excludes AI training, or just removes the word “train” and leaves room for it elsewhere.
- Licensing talk. A consultation with industry groups could turn into a negotiation, and that could end with Meta paying for access the way some AI rivals do.
- Copycat clauses. Other platforms are watching. If Meta’s approach had gone through, expect similar terms to show up across ad networks and business tools.
- Washington. With senators from both parties already on record, this could feed into wider bills on AI training data and creator rights.
For now, publishers win a short-term reprieve. The bigger fight over how AI companies get training data is far from over. Meta just learned it can’t settle it through fine print. You can find more details in the original report from The Verge AI.