Meta drops Muse Image inside Meta AI

Meta just launched Muse Image, its first image generation model out of Meta Superintelligence Labs, but the rollout hit a massive speed bump almost immediately.

Rolled out on July 7, the new model — internally code-named Mango — powers creative tools across Meta AI, Instagram Stories, and WhatsApp. It is designed to turn casual ideas into visuals, edit existing photos, and even redesign your home.

However, a particularly invasive feature triggered instant user backlash, forcing Meta to quietly kill it just three days later.

Beyond text-to-image: room redesigns and on-photo markups

Unlike image tools that focus strictly on generating art from scratch, Muse Image is clearly tuned for everyday social apps.

Users can upload their own photos to apply one-tap presets, turning themselves into claymation figures, 16-bit video game characters, or restoring old family portraits. It can also erase photobombers cleanly and generate legible in-image text for guides and readable QR codes.

One of the most practical additions is interior redesign. You take a photo of your living room, and Meta AI reimagines the space using real furniture pulled from the web or Facebook Marketplace.

Under the hood, Muse Image pairs up with Muse Spark, breaking down the prompt into multiple reasoning steps to plan layout and gather real-time web context. Editing is equally conversational: you can tap an on-screen markup tool to circle or sketch over the photo, and Meta AI tweaks that exact spot while keeping conversation memory intact.

Meta says the tool is free for everyday use, with limits tied to subscription plans. Muse Video is also already in the works.

The @-mention backlash and a three-day retreat

The feature that set off alarms was a photo-tagging integration with Instagram. Initially, Meta allowed users to @-mention any public Instagram account in their prompt, pulling that person's public photos into a newly generated AI image.

Even worse, Meta's policy stated that users would never be notified when someone used their pictures to make AI art. Opting out required digging through account settings to manually disable the feature.

Critics jumped on the move immediately. One X user quoted by TechCrunch called pulling real people into AI images without consent a privacy landmine waiting to detonate. Outlets quickly brought up Meta’s $5 billion FTC settlement over Cambridge Analytica and the 2021 shutdown of its facial recognition system.

By July 10, Meta surrendered. An official update stated the feature missed the mark and was officially pulled from the platform.

Why it matters

For regular consumers and small creators, Muse Image is genuinely useful right now. Features like circling an object on your screen to edit it or swapping secondhand furniture straight into your room photo solve real problems without requiring prompt-engineering skills.

Strategically, though, Meta showed its worst instincts before common sense prevailed. We think Meta tried to weaponize its biggest competitive moat — massive troves of personal social data — to leapfrog Midjourney and Adobe Firefly. But treating public Instagram posts as fair game for AI remixing was tone-deaf and legally reckless.

Killing the feature within 72 hours was the right move, but it leaves an important lesson for the entire industry. As AI image models shift from generating fantasy landscapes to remixing our personal lives, respecting user consent is no longer just a legal footnote — it will dictate whether mainstream users trust these tools at all.