Meta has quietly shipped the first real product out of its Superintelligence Labs division, and it is not a research paper or a benchmark chart. It’s an image generator called Muse Image, now wired into the Meta AI app, Instagram, and WhatsApp, with Facebook and Messenger to follow. The pitch is broad: advertising mockups, room decorating, creator tooling, Stories effects. The controversy is narrower and sharper — you can feed it another person’s Instagram account as a prompt.

The feature that reframes everything

Strip away the use-case slideshow and one capability defines this launch: an account handle is now an input. Type in someone’s username and Muse Image can render them into a scene they never appeared in. Meta is framing this as convenience — pull a friend into a group photo, remix a creator’s look. But the same mechanic is a stranger-targeting tool with the friction removed. You don’t need to save someone’s photos, find a reference image, or know how to prompt around a likeness. The product does the identity resolution for you.

That’s why the pushback arrived on day one rather than after some future misuse. Users aren’t reacting to a hypothetical; they’re reacting to the default. Meta’s own platforms are the training and targeting surface, and consent was never part of the flow.

Superintelligence Labs’ revealing first move

There’s a strategic tell here worth sitting with. Meta stood up a division with a grandiose name and enormous compute, and its debut ship is a consumer image toy optimized for engagement across the apps Meta already owns. That tells you what “superintelligence” means inside Meta right now: not a moonshot, but a faster on-ramp to feeding the ad and content machine. Muse Image isn’t research escaping the lab — it’s the lab justifying its budget by plugging directly into the products that pay for it.

It also explains the account-as-prompt design. A generator that only works on your own photos is a feature. A generator that works on anyone in Meta’s social graph is a network effect — every user becomes both a creator and potential subject, which is exactly the kind of viral loop Meta’s distribution is built to exploit.

The usual defense — moderation, watermarks, reporting tools — treats harm as an edge case to patch. But when the headline capability is “turn a person into a prompt,” misuse isn’t the exception; it’s the feature working as specified. Watermarking an image doesn’t restore the consent that was never asked for, and reactive reporting puts the burden on the person who got rendered.

The real question for regulators and users isn’t whether Muse Image is good at its job. By the reporting, it is. The question is whether “anyone in the graph is fair game” should ever have been the launch default — and whether Meta shipped it that way because it was the best design, or simply the most engaging one.