Meta has pulled a feature of Muse Image, its in-house AI image system built by Meta Superintelligence Labs, that let people generate images by @-mentioning public Instagram accounts they wanted to reference. The removal came after days of backlash, and Meta was blunt about why, saying it had heard the feedback that the feature missed the mark, so it is no longer available. The company launched Muse Image only days earlier, so pulling one of its headline capabilities this quickly is a real reversal.
The objection was not about how good the images looked, it was about consent. The feature was opt-out rather than opt-in, meaning it was switched on by default for adult users with public accounts, who then had to go and disable it if they did not want their images used as references by other people's prompts. Critics made a simple argument, a tool that generates images based on real people's likenesses should require those people to explicitly opt in, not leave them to discover that a setting affecting their own photos was already turned on without their say.
The pushback was fast and, importantly, organized. SAG-AFTRA, the actors' union, urged its members and Instagram users to opt out, and framed the design as a basic misread, saying that anything other than a clear and conspicuous opt-in for AI use of people's images was an utter miscalculation of public sentiment. Individual creators piled on too. Emmy winning actor Hannah Einbinder criticized the feature on Instagram, saying it had been turned on automatically, and told her followers to disable it. Within days the criticism was loud enough that Meta backed down.
What makes this more than a product footnote is how narrow the failure was. Muse Image itself is Meta's serious in-house bet on image generation, and the model was not the issue, a single policy choice was. By defaulting the reference feature to on for public accounts, Meta turned a capability people might have accepted as opt-in into something that felt like it was taking their likeness without asking. For a company with Meta's resources and legal review, having to reverse course in days over that one decision is a striking outcome.
The reason it matters beyond Instagram is that opt-in versus opt-out is one of the sharpest recurring fault lines in generative AI, and this is a clear reading of where it is landing. A default-on approach to real people's likenesses drew immediate, organized resistance, from a union with genuine leverage and from individual creators with large audiences, and even Meta chose retreat over a fight. The lesson other companies will take is uncomfortable but plain, on likeness and consent, shipping first and letting people opt out later is getting harder to defend, and users increasingly expect to be asked before their images become raw material.