Meta says it’ll probably delay its controversial new business terms again
Oct 28, 2026, 9:15 AM · The Verge

Meta is slowing down terms that publishers feared would open their sites to AI training. Backlash worked this time, but the underlying question of consent is still unanswered.
Why it matters
Meta's updated business terms were set to take effect October 30, then got pushed to November 20 after publishers objected that they appeared to let Meta train AI on news sites using Facebook and Instagram. Now Meta communications head Andy Stone says the consultation with industry groups could run past November 20, in which case implementation will be delayed again, and existing terms stay in place until then.
Meta already removed language saying it could "train" on content from those sites, but legal experts told Nieman Lab that the revision still doesn't rule it out. The pushback has come from the News/Media Alliance and from senators in both parties, including Marsha Blackburn and Adam Schiff.
From the desk
We think this is the right outcome for now, and it's a useful reminder that pressure works. Publishers, a trade group, and lawmakers from both parties pushed on the same point, and a platform the size of Meta blinked twice.
But a delay isn't a fix. The core problem is that the terms were written ambiguously enough that experts still can't say whether AI training is allowed. When a company removes the word 'train' but keeps language that might permit the same thing, publishers are right to read that as a placeholder, not a promise. Clear consent should be the floor: if Meta wants to train on publisher content, it should say so and negotiate for it.
We're generally on the side of AI being trained on broad, high-quality material. Models get better and more useful that way. But that only lasts if the people producing the material have a say and a share. News organizations depend on platforms like Facebook and Instagram for reach, which makes 'accept our terms or lose distribution' a lopsided deal. If that pattern becomes the norm across platforms, AI training rights get extracted through fine print rather than bargained for, and the economics of journalism get weaker still.
Our read: Meta's line about consulting industry groups is welcome, but the test is what the final terms say in writing. I'm watching for explicit language either excluding AI training or tying it to compensation.
Context
The terms in question govern businesses, including news publishers, that use Meta's platforms and ads tools. The concern centers on whether content from those businesses' own websites could be used to build Meta's AI products. Meta has said its existing terms remain in effect until the new ones are implemented.
Who feels it
- Publishers
- The delay buys time to negotiate, but ambiguity about AI training remains until Meta publishes final language.
- Meta
- Continued backlash from both parties raises the political cost of loosely worded AI data terms.
- Small businesses
- Any business using Meta's ad and business tools is covered by these terms, so the final wording reaches well beyond newsrooms.
- Lawmakers
- Bipartisan objections make this a likely reference point in broader debates over AI training and content rights.
What to watch
- Whether Meta formally announces a new implementation date beyond November 20
- Final wording on whether publisher content can be used for AI training
- Any licensing or compensation offers to news organizations
- Further statements or action from Senators Blackburn and Schiff or the News/Media Alliance
Companies: Meta