Introduction
This isn't a scare piece. It's a practical one. The point of documenting these clauses isn't to convince anyone to quit every platform. It's about naming exactly what a platform can and can't do with your content. And where the real pressure points are, if you decide to use them.
Here's the starting fact. These are user-generated platforms. They don't create the value themselves. The people posting to them do. That dependency is real leverage. But it only works if it's aimed at the actual mechanism, not a vague sense that something feels off.
The leverage isn't a settings toggle or a support ticket. It's the choice of what you keep supplying, and to whom.
Platform by Platform: What Exists Today
Here's what you can turn off, what you can't, and what each control actually does.
Instagram & MetaMuse Image
Meta's Muse Image launched in early July 2026. Anyone can tag a public Instagram account in a prompt and get a new AI image, built from that account's real photos. No permission asked. No notification sent. It only needs three to five clear face photos. It shipped turned on by default.
- Open your Instagram profile.
- Tap the menu, top right.
- Under “How others can interact with you,” tap Sharing and reuse.
- Turn off the toggles under “Allow people to reuse your content on Instagram and with AI features at Meta.”
Check this screen carefully. Reported toggles include Posts and Reels. But some accounts also show options for comments, captions, and mentions. Don't assume there are only two toggles. Check the full list.
You can also turn off reuse for a single post. Open the post, tap the three dots, tap Turn off reuse.
Opting out only stops future generation. Anything already made using your photos before you opted out stays in circulation. Meta has confirmed it won't be removed. The only setting that closes the door more completely is switching the account to private.
Instagram & FacebookGeneral AI Training
This is the broader use. Your posts, photos, and captions being used to train Meta's AI models generally, not just Muse Image.
- Go to Settings and Privacy, then Privacy Centre.
- Look for the notice about Meta using public information to train AI.
- Select object, give your email and reason, and submit.
This is legally binding for EU and UK users under GDPR. Meta must honor it. For US users, there's no equal legal requirement. Meta isn't obligated to act on the objection. Submit it anyway. It creates a documented record, which matters, even without a guarantee.
YouTubeNo Consumer Opt-Out Exists
There's no settings toggle for AI training use on YouTube right now. Google's legal position, the one being tested in the Lyria 3 lawsuit, is that the original terms of service, dating to 2019, already cover this use. Even though the terms never named AI or training as a use.
There's no account setting that reaches this. The pending lawsuit, public pressure, and for EU users, a GDPR-based request, are the only real levers.
TikTokNo Opt-Out for Public Content, By Design
TikTok's terms are explicit. Public content can be used to train and test its AI models. There's no opt-out. Switching a video to private afterward doesn't pull it back out of a model already trained on it. The only real control is deciding not to post publicly at all.
Recording platformsRiverside and Similar Tools
Covered in an earlier piece: opt-out from AI training exists only on Business-tier plans, through a written agreement. Pro and Standard users don't have a settings-level opt-out. The available leverage is a direct written question, the same kind sent to Riverside in this series.
Where the Actual Leverage Is
None of these platforms exist without content people keep supplying. That's not a slogan. It's the actual mechanism the leverage runs on. Two forms of it are available to any creator right now, and they're stronger than anything a settings menu offers.
Lever 1Withhold new content
Deleting your archive rarely undoes anything. TikTok and YouTube are both explicit that removing a post doesn't pull it back out of a model already trained on it. But a model only stays useful if it keeps getting new material. Continuing to post is what keeps the platform's terms working in its favor. Not posting is the one thing a settings page can't override.
This doesn't require an all-or-nothing boycott. Pausing new uploads to one platform, or posting somewhere else first, is a real choice. It's a lever this series can point to with no caveat about whether the platform has to honor it. It doesn't ask permission.
Lever 2Go independent
This has gotten more realistic, not less. Podcasting already had an independent option built in from the start. An RSS feed isn't owned by any single platform. Hosting your own feed, instead of depending on one platform's app, was the original model.
Self-hosting tools have matured too. Ghost, for example, is fully self-hostable. Creators who run it themselves control the code, the data, and the design outright. No platform clause reaches content that never sat on that platform's servers in the first place.
Independence shifts the burden, it doesn't remove it. Self-hosting usually means no built-in discovery or algorithm-driven audience growth. Security and maintenance become the creator's job. That's a real cost, not a free win.
The fallback, not the leverDocumented objections
Filing a GDPR objection, or sending a written question like the one sent to Riverside, still matters. But as evidence, not as force. Neither one is the thing that makes a platform change course on its own. They're the record that supports someone else's case later, whether that's a journalist, a regulator, or a future lawsuit.
Two examples show documented pressure can eventually lead to real change, just not quickly. Meta shut down Facebook's facial recognition system in 2021, after lawsuits and regulatory pressure. Muse Image's opt-out-by-default design got fast, visible backlash, including a public statement from Creative Artists Agency. Companies with real reputational risk do respond to sustained, specific pressure. But the pressure that moved them was public and collective. Not one person's settings change.
What Actually Moves a Platform
Specific requests beat general complaints. “Give people an opt-in, not opt-out, for AI training” is a demand a company can act on. “This feels invasive” isn't.
Collective pressure has more weight than individual settings changes. One person opting out changes nothing about a platform's policy. A visible, sustained, specific ask from the whole community does.
Withholding content and building independent options are levers a platform can't simply decline to honor. Unlike an opt-out request, they don't need the platform's cooperation to work at all.
Where This Fits
This is one piece in a larger series about Retroactive Consent, the pattern of old agreements being stretched to cover new AI uses. The goal here isn't to get anyone to delete their accounts. It's to make sure the choice to stay is actually informed, and that leaving, for one platform or one type of content, is treated as the real structural lever it is. Not a last resort.
Where This Goes Next
Retroactive Consent isn't limited to platform settings. The same structure — old terms, stretched to cover a use never named — shows up anywhere a platform, a vendor, or a client-recording tool sits between a person and their own voice, face, or written work. This series will keep applying that same pattern, case by case, to:
Read the rest of the series: How Old Terms Cover New AI Uses · Riverside AI Training · YouTube, TikTok & Meta: What Their AI Terms Allow