Meta's new terms turn publishers into *free AI training camps*

The new terms in plain language

Meta's updated terms of service, set to take effect on 30 October 2024, include a clause that grants the company the right to use content posted by publishers for training its AI systems ^[1]. The terms apply to anyone who continues using Facebook and Instagram after that date. For publishers, that means every article preview, image and post becomes potential training material.

The mechanics are simple. Stay on the platform past 30 October and you agree to the new terms. Post content after that date and Meta can use it to train its models. It's a bit like putting your newspaper on the staffroom table and coming back to find someone has photocopied it for a training manual. Only in this case, the manual teaches AI how to write articles.

Press Gazette quotes publishers calling the move a 'shameless land grab' ^[1]. The complaint is less about AI training in general and more about the lack of negotiation. Publishers built their Facebook pages over years, and now Meta is changing the deal: stay and we take your work, or leave and lose your audience.

What publishers face

The immediate question is reach. Facebook and Instagram remain major traffic sources for many publishers. Leaving means giving that up. Staying means handing over content that might, down the line, train a system that competes with journalism itself.

The terms don't specify how Meta will use the content, which models it will train or whether publishers will know when their work has been used. For a publisher weighing the decision, that uncertainty matters. You're not just agreeing to today's AI tools. You're agreeing to whatever Meta builds next.

Some publishers have the scale to negotiate directly with Meta or to treat Facebook as optional. Smaller outlets don't. For them, the choice is between two uncomfortable options: accept terms they dislike or abandon a platform their readers use.

The evidence question

If a publisher later disputes how Meta used its content, the question will be what was posted and when. Meta's terms apply to content posted after 30 October. Proving what appeared on your page on a given day, and that you didn't silently remove it to avoid the new terms, isn't straightforward once the post is gone.

Screenshots can be edited. Web archives don't capture everything. The publisher's own records may be questioned, especially if Meta's and the publisher's versions differ. That's the gap between 'we posted this' and 'we can show we posted this, unchanged, on that date'.

How a digital witness fits here

This is a case where provability matters from the start. A publisher who seals each post as it goes live has a record Meta can't dispute. The thing to seal is the post content, the publication timestamp and the page it appeared on. Sealing takes a fingerprint of that data, not the data itself, so the posts stay on the publisher's systems.

When the fingerprint is sent to WitniumChain, it gets a permanent, timestamped entry. If a dispute arises, the publisher shares the original post and the witness record. Anyone can recalculate the fingerprint and check that it matches the sealed version. Change one word in the post and the fingerprint won't match.

What that shows is narrow but useful: this exact post existed at this time, and it hasn't changed since the seal. It doesn't show whether the post was lawful, whether Meta actually used it or whether the publisher's departure from the platform was timely. It just closes the door on one line of argument, the one that says 'you can't prove what you posted'.

What it doesn't do is stop Meta from using the content under the terms. If you posted after 30 October and stayed on the platform, the terms apply whether you sealed the posts or not. The witness is for later, when the question is what happened and when.

Starting is straightforward: send a fingerprint with the REST API, the SDK or the MCP server each time a post goes live. The content never leaves your systems. The first 2,000 witnesses a month are free.

How it would work

1. The publisher's CMS sends a fingerprint of each Facebook post (content, timestamp, page URL) to WitniumChain at publication.
2. WitniumChain returns a witness ID. The publisher stores the ID alongside the post.
3. If a dispute arises, the publisher shares the original post and the witness record. Anyone recalculates the fingerprint and checks it against the sealed version.

What this means for you

If you publish on Facebook or Instagram, decide before 30 October whether you're staying. Staying means accepting the new terms. If you stay, consider sealing each post as it goes live. That way, if the question is ever what you posted and when, you have an answer that doesn't rely on Meta's records or on screenshots.

If you leave, seal a record of when you stopped posting and, if possible, what your last posts were. That timeline may matter if Meta later claims content was posted after you say you left.

For publishers who negotiate directly with Meta, the same principle applies: seal the agreement and each communication. If the terms are later disputed, the record of what was agreed, and when, is yours.

What to watch

The 30 October deadline is firm ^[1]. After that, the new terms apply to everyone who stays. Whether other platforms follow Meta's approach is an open question. If they do, the evidence question becomes industry-wide: how do you prove what content you posted, when you posted it and when you stopped?

Sources

1. Press Gazette, 'New Facebook terms for publishers "shameless land grab"', 9 October 2024, https://pressgazette.co.uk/news/facebook-terms-publishers/