Meta launched Muse Image on Tuesday, a new AI image-generation tool built into Instagram, Meta AI, and WhatsApp that lets users create original images, edit existing photos, and generate custom ads. The feature that’s drawn the most criticism: anyone can tag a public Instagram account by username and use that account’s photos as reference material for an AI-generated image, without the account owner’s knowledge or consent, and every public Instagram profile was automatically opted into being used this way when the feature launched. Only private accounts and users under 18 are automatically excluded. YourNewsClub pins the automatic opt-in, rather than the image-generation technology itself, as the actual design decision driving the backlash: Meta could have launched the identical capability as an opt-in feature and let interested users enable it, and chose the opposite default for everyone with a public account instead.
The consent gap compounds in a specific way: Instagram doesn’t notify a user when someone else generates an image using their tagged public account, meaning there’s currently no way to know how many times, or in what context, your own photos have already been used as AI reference material since the feature launched. Opting out only prevents future use – any images already generated using an account’s photos remain in circulation regardless of a later opt-out, and Meta has no mechanism to retroactively identify or remove them.
To opt out, users open their Instagram profile, tap the menu icon, go to “Sharing and reuse,” and turn off the toggles for “Allow people to reuse your content on Instagram and with AI features at Meta” for both posts and reels – settings that, notably, aren’t part of Instagram’s main privacy menu and require a person to already know the feature exists and go looking for the controls specifically. YourNewsClub maps the setting’s location, buried under a secondary menu rather than surfaced during onboarding or in the primary privacy settings, as consistent with the same default-collection pattern showing up across the industry this year, where a technically-available opt-out exists but isn’t placed anywhere a typical user would encounter it without already knowing to search for it.
Maya Renn, whose work focuses on the ethics of computation and access to power through technology, frames the impersonation risk directly: “Public Instagram photos were already being harvested for deepfakes and identity-verification fraud before this feature existed. Giving anyone an official, built-in tool to generate AI images referencing a specific person’s public account, without notifying that person it happened, doesn’t just fail to solve that problem, it hands the same capability a corporate stamp of legitimacy and removes the small amount of technical friction that used to separate casual misuse from more sophisticated fraud.” Owen Radner, who models digital infrastructure as energy-information transport systems, places the default-architecture angle: “The technical decision to auto-enroll every public account, rather than requiring active opt-in, wasn’t incidental – reaching meaningful training and reference data at launch scale requires exactly this kind of default. The privacy controls exist, but they were clearly designed after the product decision to maximize participation was already made, not before it.”
Meta’s approach mirrors a broader pattern in how it has handled AI training consent for European users, where the company relies on GDPR’s “legitimate interests” legal basis rather than requiring opt-in consent – a position a European privacy advocacy group has formally challenged. YourNewsClub logs that European legal challenge as the more consequential process to watch relative to Muse Image specifically: if regulators there successfully narrow what “legitimate interests” can justify for AI training and reuse, it would constrain the same default-opt-in logic Meta just applied to Muse Image globally, not only the European rollout of that specific feature.
Your News Club rates the retroactive-removal gap – the fact that opting out doesn’t delete AI images already generated using a person’s account – as a more significant unresolved problem than the opt-out process itself: a person can eventually find and disable the setting, but there’s currently no path, technical or otherwise, to learn what’s already been generated using their likeness or to have it removed.