Meta quietly kills Muse's AI photo generator after backlash

Meta yanked a feature from its new Muse Image model this week, and honestly, it’s hard to see how anyone thought it was a good idea to begin with. The tool let you generate AI images of any public Instagram user simply by @-mentioning their handle. No consent required. No notification. Just type a username and the model would spit out a synthetic version of that person. It was on by default. To escape it, you had to dig into Instagram’s settings and manually opt out.

The backlash was immediate and predictable. Critics pointed out the obvious: giving anyone the power to generate images of someone else without permission is a privacy nightmare dressed up as a creative feature. Meta, to its credit, didn’t try to tough it out. The company admitted the feature “missed the mark” and killed it just days after the announcement. They claimed the original intent was to offer control — letting people decide if their public content could be referenced this way. But making it opt-out rather than opt-in betrayed that logic entirely.

It’s not hard to trace where the inspiration came from. OpenAI’s Sora app briefly offered a “cameo” feature that let users create AI versions of themselves and, crucially, grant explicit permission for others to use those cameos in videos. It was a viral moment that faded almost as fast as it appeared, but at least the consent architecture was built in from the start. Meta seemed to borrow the fun part and forget the permission part.

Europe would have been a brick wall for this feature anyway. GDPR doesn’t look kindly on scraping public profiles to train generative models, especially without affirmative consent. The regulatory headaches alone might have forced Meta’s hand even without the public outcry. But the speed of the reversal suggests something else: even Meta’s product teams know the Overton window on AI image generation is shifting fast. What felt like a clever demo six months ago now lands as a liability. The question isn’t whether Meta tries something like this again — it’s whether the next attempt comes with the guardrails already bolted on, or if we’re just waiting for another apology.