> Quick summary — 3 key points about Meta’s Muse Image model
- Pulls real people into AI-generated images: This feature lets users bring other people’s faces from Instagram (even without tagging or asking permission) into images the AI generates itself, pulling facial data from public profiles to process.
- The problem: consent disappears in the middle: The person in the photo doesn’t know they’re being used, never opted in, and has no way to opt out beforehand — unlike other AI features that typically ask permission before using personal data.
- Risk of becoming a social deepfake tool: Because the input is real photos of real people, the risk of fake images, misuse, or out-of-context images is higher than with typical AI generators that rely on synthetic datasets.
Note: There’s no confirmed user count or incident-report data available yet. We’ll need to wait for Meta to clarify further.
What this feature actually looks like in use
Picture this: you open the app, type a prompt like “me and my friend at the beach,” and Muse doesn’t generate a stranger’s face — it pulls in the face of a friend you actually follow and drops them right into the scene.
The “aha” moment here is that this isn’t just generic AI image editing — it knows what “this specific friend” looks like from their Instagram data, then blends them into a scene that never actually happened.
In short, the system shifts from “generate a fictional person” to “use a real person you know” — a line that sets it apart from every other AI photo generator we’ve gotten used to so far.
The day you see your own face in a photo you never took
Imagine opening Instagram and finding a friend has tagged you in a photo standing at a concert you never went to, next to someone whose face you recognize but have never actually met. The first thought is “when was this taken?” — before you realize it’s an AI-generated image built from your face, pulled from old photos on Instagram.
The problem is nobody asked first. You never consented to anything. Just having an Instagram account that a friend can tag is enough to get pulled into a scene that never actually happened.
This is exactly what makes consent the central issue throughout this piece — the line between “a friend sharing a funny photo” and “anyone can use my face” turns out to be much thinner than expected.
Where Muse sits in Meta’s AI battlefield
Meta already has AI tools scattered across multiple touchpoints — Meta AI in chat, Imagine with Meta AI focused on prompt-based image generation, and AI Studio for building bot personas. Muse is different: it didn’t launch as a separate app or standalone feature — it’s embedded directly into Instagram, wired straight into users’ real social graph.
That’s the strategic angle — Meta isn’t just competing on how good its AI is at generating images, it’s competing on “who has more social data for AI to draw on.” Pushing this feature into Instagram right now aligns with Meta’s broader push to tie AI directly to engagement on its core platform, rather than letting AI remain a separate toy the way competitors treat it.
From the old Meta AI to Muse — how much has changed?
Meta’s previous image-generation model could only generate images from generic prompts — it had no idea whose face was whose. Muse is different: it can pull data straight from Instagram to reference real people’s faces. This is the single most significant shift — not that images look better or generate faster.
| Factor | Meta AI (previous) | Muse |
|---|---|---|
| References real people from IG | Not possible | Possible |
| Data source used | Text prompts only | Instagram data/photos |
| Privacy controls/safeguards | Not an issue to begin with | A newly emerging issue |
| Risk of being pulled into an image without consent | Virtually none | Real risk |
Put simply, Muse isn’t just a model upgrade — it changes the entire “scope of data” the AI can access.
What can you actually do with pulling friends into a photo?
Picture this simple case: you want a beach-trip photo with a friend who couldn’t make it — just tag their IG handle and Muse pulls their face right into the scene. No need to wait for anyone to actually take the photo.
The fun side: create fantasy group-friend scenes, restyle images into cartoon/vintage looks, share straight to your story without opening a separate editing app.
The concerning side: the person pulled into the photo might have no idea their face ended up in an image that was never actually taken — and if it’s used to create a fake image realistic enough to be indistinguishable from reality, like placing them in a situation they were never actually in, that becomes a problem immediately.
The point is: the exact same feature can be used to create cute fake memories — or to fabricate photo evidence. The line between them is consent, plain and simple.
Compared to competitors — who else dares to let you pull other people’s faces into images like this?
Let’s line up the major AI image-generation players and see how each one answers the question: “can you pull another person’s face into an image?” The answers are quite different.
| Factor | Meta Muse | Google Gemini (Nano Banana) |
|---|---|---|
| Pulls in non-consenting people's faces | Allowed under certain conditions | Blocked by default |
| Identity-verification guardrails | Based on friend network | Focused on blocking real people's images generally |
| Risk of realistic fake images | Higher, since it can pull real faces | Lower, due to stricter restrictions |
The clearest difference is that Muse is designed to let you “pull” people you actually know into an image — that’s its social selling point. But the same thing that makes it a selling point also makes it a vulnerability. Other players chose to block this outright for safety, trading away some of the fun in the process.
Weighing the real pros and cons
Pros
- +Generates images quickly and naturally, blending seamlessly into your normal feed
- +Pulls friends or acquaintances into images using their actual likeness, making content feel more like a real story
- +Easy to use — no need for external photo-editing tools
Cons
- −A clear consent problem — people pulled into images never agreed to it or knew about it beforehand
- −Risk of fake images that are hard to distinguish from real ones
- −Even people who don't use this feature at all can still have their face pulled into someone else's image
To be blunt, this is a trade-off Meta chose deliberately — rolling out a fun feature while pushing far too much of the safety burden onto users to figure out for themselves.
A price paid in privacy, not in money
This feature is free — but the real cost isn’t coming out of our wallets. It’s being paid by other people who never even asked to be part of it.
Someone whose face gets pulled into an image has no way of knowing in advance where that image will end up or in what context. If it’s an image that damages their reputation or is embarrassing, the harm happens before they even find out.
There’s also the opt-out burden — systems like this are typically set to “on” by default, leaving users to go find the switch to turn it off themselves. Ordinary people who don’t even know this feature exists have almost no chance to protect themselves.
The deepfake risk is even more complicated — the line between real and generated images keeps getting blurrier. Society ends up shouldering the burden of proving what’s real, even though Meta is the one who opened this door.
What Meta still hasn’t answered — and what to watch next
Meta still hasn’t clarified whether it will switch the system from opt-out to opt-in, or whether it will build a mechanism to notify people when their likeness is pulled into an AI-generated photo. This is exactly the kind of thing privacy regulators in multiple countries are likely to start watching closely, since using someone’s facial data without clear consent could easily run afoul of personal data protection laws.
User pressure matters just as much — if enough people complain or start moving to other platforms, Meta may be forced to adjust its default settings faster than it originally planned.
While waiting for clarity from Meta, one thing you can do right now is check your own Instagram privacy settings — see how much control you have over whether AI can use your profile photos or images you’re tagged in. Don’t wait for a feature like this to surprise you later.