Meta pulled Muse Image in 3 days after Instagram backlash over public-post AI use

Meta launched Muse Image on Tuesday, and it was no longer available by Friday. The feature, available through Meta AI, let users generate AI images based on public Instagram posts, using the first image-generation model from Meta Superintelligence Labs. Meta had framed it as a creative tool with user control over whether public content could be referenced, but it pulled the option after the rollout missed the mark with feedback.

In a Friday update, Meta said the feature is no longer available. In an Instagram blog post, Meta explained: “Our intent was to provide a useful creative tool and to give people control over whether their public content could be referenced in this way. We've heard the feedback that this feature missed the mark, so it's no longer available.” That fast reversal is the real story here: a routine product drop turned into a consent, privacy, and deepfake backlash in about 72 hours.

To understand why this landed so hard, it helps to map how these AI features are typically marketed versus how they are experienced. The announcement was tied to Muse Image, an AI image generation model. Users could take public Instagram posts and use them as reference inputs to create new AI images. That sounds straightforward to a product team. But to users, it raised two immediate issues: privacy concerns and the fear of deepfake-like content. In the source, users quickly rebuked the feature and shared instructions for opting out, with advice circulating on platforms like X and Reddit.

Meta did not kill everything related to Muse Image. The source is explicit that other Muse Image features, like the ability to make edits directly on photos, are still available. This matters for decision-makers because it shows a common pattern in AI product management: the model can keep running, while a specific “data use” behavior gets removed when the governance story fails. In other words, the backlash was not necessarily about AI image generation in general. It was about referencing public Instagram content in a way that users and critics believed lacked adequate consent.

The criticism came from both everyday users and organized entertainment stakeholders, and that combination is where the PR risk gets expensive. Privacy advocates attacked Meta for consent and forced use. Apar Gupta, the founding director of the Internet Freedom Foundation, criticized Meta in a video posted on X on Friday. The core of the concern, as quoted in the source, was that platform ownership should not be treated as permission: “Just because Meta owns one of the largest social media platforms, and we're forced to use it, it's been taking it as an excuse to violate our consent and privacy again and again.”

Then labor and talent groups moved from commentary to policy language. SAG-AFTRA, which represents about 160,000 entertainment and media professionals, urged members to opt out. In a statement to Business Insider, SAG-AFTRA said that “anything other than a clear and conspicuous OPT-IN for these types of uses of Instagram users' images is unacceptable, and an utter miscalculation of public sentiment regarding the obvious dangers and harms inherent in such use.” The union also applauded Meta’s discontinuance, calling it responsible given “the dangers of nonconsensual digital replicas well known to all.” The Creative Artists Agency, in a statement shared with Variety, similarly argued: “No one's name, image, likeness, voice, or creative work should be used by any third party, including AI models, without clear, documented consent.”

This is also not Meta’s first rodeo with AI policy friction. The source connects the backlash to broader tensions across the industry. It notes that AI has become a source of concern for entertainers who worry about stealing likeness, voices, or phrases without consent. And it cites that celebrities including Matthew McConaughey and Jeremy Clarkson have registered trademarks to protect their likeness from AI. That is a sign of a market reacting faster than legal frameworks, using the tools available today while the rules are still catching up.

There’s another important comparison in the source: OpenAI faced similar backlash after releasing Sora 2, an AI video-generation platform, in 2025. The platform initially allowed users to create content featuring trademarked characters, which drew ire from entertainment companies such as Studio Ghibli. Even after OpenAI secured a partnership with Disney to legally use those characters, Sora 2 shut down in March. The implied lesson for executives is not “AI is always bad.” It is that when AI features touch rights and consent, the business outcome can turn abruptly, regardless of model capability or initial product intent.

Second-order implications for boards and leadership teams: this is a governance problem as much as a technology problem. Muse Image’s rollout illustrates how quickly user trust can erode when a feature is perceived as using public content in ways that should, in critics’ view, require explicit opt-in. It also shows how organized stakeholders can accelerate internal decisions by raising stakes around consent, deepfakes, and “nonconsensual digital replicas.” For leaders building or buying AI capabilities, the bar is shifting from “we meant it as creative” to “users have clear, conspicuous control over use,” and that bar is being enforced socially, commercially, and sometimes via sudden product shutdowns.