Meta abruptly removed a default setting in its Muse image generator that let the AI freely pull photos and videos from public Instagram profiles, following intense criticism from users and entertainment-industry groups. The company acted late last week, turning a once-automatic data pipeline into an opt-in feature—a rare reversal that highlights growing tension over how tech giants use public content to fuel generative AI.

What Actually Changed Inside Muse

At the center of the uproar is Muse, Meta’s generative AI tool that turns text prompts into images. Prior to the change, Muse had a baked-in capability to reference publicly shared Instagram photos and videos—essentially treating the entire library of public Instagram content as a searchable, remixable dataset. Any user with access to Muse could describe a visual idea, and the tool would silently scan public Instagram posts for matching elements, incorporating them into new AI-generated images without notifying the original creators.

The critical detail, as confirmed by Meta’s advisory, was that this referencing behavior was enabled by default. There was no toggle, no consent flow, and no way for public-profile holders to know their content had been swept into someone’s AI creation. The change, implemented on Friday, flips that default off: now, Muse no longer automatically draws on public Instagram media unless a user explicitly instructs it to do so—and even then, only under a more restricted set of rules that Meta has not fully detailed.

For users who had already become accustomed to the feature, the tool now behaves differently. Prompts that might have once pulled a celebrity’s public photo or a scenic travel shot from a stranger’s feed now return generic AI generations based solely on Muse’s training data—unless the user actively uploads an image or provides a specific link. Meta has not retroactively deleted any previously generated images, meaning content that may have incorporated public Instagram visuals before the change remains accessible to those who created it.

What This Means for You

The practical impact depends entirely on how you interact with Instagram and Meta’s AI ecosystem.

For Everyday Instagram Users

If your account is public, any photo or video you share can still be seen by anyone on—or off—the platform. But the key shift is that your content is no longer being vacuumed up by a default AI engine for image generation. Does that mean Meta won’t use your public posts for other AI purposes? Not at all. This change specifically stops the live, on-demand referencing within Muse; it does nothing to alter Meta’s broader data-scraping practices for training its underlying models, which rely on massive amounts of public content. So while you’ve been spared from having your vacation snaps randomly appear in someone’s AI-generated art, your data is still building Meta’s AI muscle behind the scenes.

For Creators and Public Figures

Photographers, artists, influencers, and anyone who relies on Instagram for professional visibility should treat this as a narrow win. The feature that could have let a fan mash up your face with a fictional character is now behind a permission wall. However, the risk hasn’t disappeared—it’s just less convenient. Anyone with a download of your public image can still feed it into any AI tool manually. Entertainment-industry groups like SAG-AFTRA and the Writers Guild have argued for broader protections, and Meta’s move here is likely a concession to avoid legal fights over right-of-publicity and copyright claims that were already brewing.

For AI Tinkerers and Power Users

If you’ve been using Muse to quickly prototype visual ideas, you’ll find the tool slightly less magical. The instantaneous access to the world’s visual culture through Instagram was a huge differentiator. Now you’ll need to source reference images yourself or craft more detailed prompts. Power users who relied on the default behavior for rapid iteration will have to adjust workflows, but the underlying AI remains capable—it just requires more deliberate input.

How We Got Here

Meta’s Muse hasn’t existed in a vacuum. The feature emerged from the company’s years-long push to dominate generative AI, using its unmatched reservoirs of user-generated content. When Muse was first introduced in late 2024, the Instagram integration was a headline feature: a way to distinguish Meta’s offering from Midjourney, DALL·E, and Stable Diffusion by tapping into the world’s largest public visual database in real time. Executives argued that public posts are effectively out in the world, and using them for AI generation was no different than a human browsing the feed for inspiration.

But the optics were terrible. Users began reporting examples where the AI apparently reproduced distinctive elements from strangers’ photos—a tattoo design, a unique piece of furniture, an identifiable face—without attribution. Shortly after, major entertainment unions and industry groups sent a joint letter to Meta, warning that the tool opened the door to mass-scale impersonation and copyright infringement. Social media erupted with a familiar call: “If you’re not paying for it, you’re the product.”

Meta initially defended the feature, pointing to its terms of service that grant the company broad licenses to use public content. But the backlash grew louder when privacy advocates demonstrated how easily the tool could create photorealistic images of real people by pulling from their public Instagram profiles. The pressure intensified until Meta’s abrupt Friday fix—a move that surprised few observers but was executed with little fanfare, suggesting the company wanted to bury the reversal.

What to Do Now

For most people, the immediate steps are simple but important:

  • Check your Instagram privacy settings. If you absolutely don’t want your photos or videos to be used in any AI training or generation, the safest option is to make your account private. Public profiles remain eligible for Meta’s broader AI training and could still be referenced if a user manually provides your image to Muse.
  • Review Meta’s data use policy. In Instagram’s settings, under “Your activity and permissions,” look for “Data management” options. Meta offers limited controls for how your information is used for AI improvements, though these vary by region and are often opt-out rather than opt-in.
  • Watermark or downscale public images. While no defense is foolproof, creators who want to deter AI ingestion can add subtle watermarks or reduce image resolution, making automatic referencing less likely to pull that content.
  • Pressure for transparency. If you rely on Instagram for exposure, push Meta to provide a content attribution tool or a registry where creators can see if their work has been used in AI generations. Currently, no such mechanism exists.

Outlook

The Muse reversal is a signal, not a settlement. Meta will almost certainly refine how its AI tools interact with user content, but the fundamental battle over public data isn’t ending. Earlier this month, the company quietly updated its privacy policy to reaffirm that public posts may be used for AI training globally, with limited opt-outs only in the EU and a few other jurisdictions. Entertainment-industry groups plan to continue lobbying for federal right-of-publicity legislation that would force explicit consent for AI use of individual likenesses, while platform users are left navigating a landscape where every post is a potential AI ingredient.

Keep an eye on Meta’s next AI product launch. Any new creative tool that ships with a “default-on” content referencing feature will face instant scrutiny, and the company’s willingness to walk back this one shows that user and industry pressure can still force a pivot—at least for now.