Meta turns public Instagram photos into raw material for Muse Image

Meta is rolling out Muse Image, according to The Verge, the first AI image generation model built by its Superintelligence Labs division. The model now powers image tools in the Meta AI app, Instagram and WhatsApp, with Facebook and Messenger planned next.

Muse Image brings generation into the social graph

The most sensitive feature is @ mention support in prompts. A user can mention another Instagram account, and Meta says Muse Image can use public photos to build a visual with that person. Meta also says users can control how others reuse their content for AI.

The model can transform images, suggest prompts for invitations and postcards, edit photos through direct drawing and redesign rooms from an image pulled from Facebook Marketplace or elsewhere on the web. Instagram Stories in the US are getting 30 new AI effects before a broader rollout to other countries and more Meta apps.

Generative AI is moving from an editor into the reputation layer

With a normal image model, the problem is the prompt and the output. With Meta, the problem also includes the relationship between the person writing the prompt, the person being generated and the person who absorbs the social consequences. Instagram is not a blank canvas. It is a database of relationships, faces and public signals.

That is strategically powerful for Meta. It has distribution, photos, identity and messaging inside one product. For users, it is a different bargain from a standalone image generator: their public profile can become input for someone else’s creation.

Controls matter more than the agentic label

Alexandr Wang describes Muse Image as agentic because it works with Muse Spark to reason through the prompt, search the web and plan before generation. That may be technically interesting, but the public test is simpler: whether likeness controls are clear, timely and enforceable.

The Verge does not spell out a detailed opt-in or opt-out flow, so caution is warranted. Public photos are not the same thing as social consent to appear in someone else’s AI image, especially when the result can go back into feeds, Stories or chats.

The first failure will come from an ordinary prompt, not a benchmark

The key signal is how Meta sets defaults, notifications and boundaries around sensitive uses of likeness. Users need to see clearly who can invoke them, where to disable it and what happens when the feature is abused.

If the controls live deep in settings, Muse Image will not be merely a creative feature. It will be a test of whether Meta can distinguish public content from public permission.

Lilith's verdict

Meta has put a photo booth in the middle of the square and handed strangers other people’s portraits as props. The fun stops when someone recognizes themselves in an image they never agreed to sign.