‘Built for your world’ or a privacy trap? How safe is Meta’s new Muse Image AI model?

Meta has officially stepped up its artificial intelligence game by rolling out its most advanced media generation model yet -Muse Image. Developed by Meta Superintelligence Labs, this new tool isn't a standalone website or app. Instead, Meta is embedding it directly into the platforms billions of people already use every day—Instagram, WhatsApp, Facebook and Messenger.

While the technology offers impressive, seamless creative features, it has instantly run into a wall of sharp criticism. A controversial feature that allows the tool to pull and manipulate photos from public Instagram accounts has raised massive red flags regarding data privacy, consent and identity theft.

How Muse Image changes everyday posting

Rather than treating AI as a gimmick where you simply type a prompt and get a static picture, Meta is weaving Muse Image into the actual workflow of social media and advertising.

• Social and chat integration: On Instagram, the model already powers more than 30 unique Stories effects. Over on WhatsApp, users can generate images directly inside their chat windows.
• Smart editing, not just creation: Muse Image acts like a lightweight editor. You can blend multiple photos, apply presets and revise specific sections by sketching or annotating on the screen. Meta says it can handle precision commands—like clearing up a blanket of fog in a landscape photo to reveal a valley below or changing the petals of a flower into a rainbow gradient.
• Invisible watermarks: To address deepfake concerns, Meta has introduced a ‘Content Seal’. This invisible watermarking system embeds a hidden provenance signal into the file. Meta claims this digital seal remains intact even if the image is cropped, compressed, resized or screenshotted. A public detection tool is also being previewed so users can verify if an image was built using Meta AI.

The big catch: Automated image-scraping

The most sensitive and heavily criticised aspect of the rollout is the Instagram @Mention feature.

With this tool, any user can type a public Instagram username directly into an AI prompt. Muse Image will then automatically claw photos from that public account to influence or generate a completely new, fake picture of that person—all without asking for explicit permission first.

The opt-out trap: This feature is automatically turned on for all public accounts worldwide. Unless a user actively dives deep into their privacy settings and manually toggles it off, their public likeness is fair game for anyone else's AI prompts.

While Meta insists that users have control over content tagging for AI creation, critics argue that this safeguard is far too narrow. Instead of requiring a fresh, explicit approval every time someone's likeness is invoked, the burden is placed entirely on the user to find and disable the setting.

High stakes for creators and businesses

This automated scraping presents major risks, particularly for influencers, creators and public figures whose faces represent their commercial livelihood. The ease with which lookalike images can be generated introduces massive potential for impersonation, scams, and targeted misuse.

There are legal landmines for commercial users too. If a business leverages Meta's Advantage+ ad stack to auto-generate campaign assets, and the AI accidentally pulls a real person's face into an advertisement, the liability shifts. Legally, the business running the ad—and not Meta—would likely face the brunt of privacy lawsuits and public backlash.

While industry experts don't predict a mass exodus from social media platforms just yet, this update is expected to make everyday users "far more selective" about the personal photos they choose to share publicly in the future.