Meta deactivated Muse Image, an Instagram feature that generated AI pictures of public accounts when you tagged them. The company updated its announcement to say the feature "missed the mark" and is no longer available.

Muse Image launched just days earlier as part of Meta's Muse family of AI tools. The feature used spatial reasoning and visual memory — technical ways of saying it understood composition, context, and visual patterns — to generate images based on an account's posted content. In practice, anyone could tag a public Instagram profile and create AI-generated deepfakes of that person, all drawn from their existing posts.

The catch: the feature was on by default. If you didn't want your content used this way, you had to dig into Settings and manually flip a switch labeled "Allow people to create with and reuse your content," or make your profile private. For the millions of public accounts on Instagram — creators, celebrities, small businesses — not doing anything meant your content could be used to generate images of you.

Creative Artists Agency, which represents talent in entertainment, contacted Meta directly about the feature. CAA's position was straightforward: no one's likeness, name, image, or creative work should feed an AI model without explicit, documented permission. The Screen Actors Guild union (SAG-AFTRA) told its members to opt out while the feature was live. Within roughly a week of launch, Meta killed it entirely.

Meta hasn't explained what "missed the mark" means in detail. The company hasn't said whether a revised version might return, whether the underlying AI model will be used elsewhere, or what specifically changed between the July 8 launch and the deactivation this week.

The speed of this reversal is unusual. In past controversies over voice cloning and AI-generated likenesses — debates that typically stretched over months — we saw studios and unions push back, then watch the industry slowly recalibrate. Here, the entire cycle compressed into days: launch, pressure from talent agencies, union guidance, shutdown.

The opt-out-by-default design is what matters most here. You see this pattern everywhere: in how apps share your data, target ads to you, and now in how generative AI tools ask permission. The friction of saying no falls on you, the user. It's a design choice, not an accident. Meta's updated language doesn't clearly say whether the default setting itself was the mistake or whether the whole feature was the problem from the start.

What CAA asked for — permission that is "clear" and "documented" — sets a much higher bar than a buried menu setting. Whether that becomes the standard for any product that lets one person generate synthetic media of another person is still an open question. Here's what's worth flagging: entertainment unions and talent agencies have real leverage. Meta responded to institutional pressure from CAA and SAG-AFTRA, not to individual users figuring out how to opt out of a setting they may never have seen.

This pattern recurs often in technology. A capability ships broadly, marketed for its creative potential, before anyone has fully worked through the consent framework around it. Muse Image's technical capabilities were likely designed for many legitimate uses — but a feature that turns any public account into training material for deepfakes, switched on by default, was always going to bump against basic questions about what "public" means when it comes to consent.

Meta has given no timeline for a revised version, no details on what protections might govern generative features tied to real people on its platforms, and no account of how this will shape future product launches.