Meta’s Superintelligence Labs, Mark Zuckerberg’s flagship initiative to build “a personal superintelligence that knows us deeply”, launched its first public product on July 8 with a name that, in hindsight, was setting expectations far too high. Muse Image, an AI-powered image generation model with 30 filters designed to work with Instagram content, lasted less than 72 hours before being pulled from the platform.
The feature that caused the backlash was straightforward in design but catastrophic in its privacy implications. Muse Image allowed users to @mention any public Instagram account in Meta AI and “generate creative AI images featuring them”, applying stylised filters to someone else’s content by default, without requiring that person’s opt-in. Meta’s launch announcement described it as making AI “more personal, fun, and social.”
The response was immediate. SAG-AFTRA, the actors’ union, called the feature “an utter miscalculation of public sentiment regarding the obvious dangers and harms inherent in such use.” The union demanded a “clear and conspicuous opt-in” for any use of members’ images. Privacy advocates pointed out that Meta’s track record, three times cited in the article as “leading the world” in social network scale, online abuse, and community backlash over privacy, should have made the outcome predictable.
By July 11, the feature was gone. Meta’s retraction statement read: “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.”
The episode is an awkward debut for Meta’s Superintelligence Labs, which Zuckerberg has positioned as the company’s long-term bet on AGI-level AI. Some of the Muse Image filters had been designed in collaboration with Instagram creators, a deliberate strategy of prioritising creator content over traditional media, which only amplified the backlash when those creators saw how their work could be repurposed.
The Register’s report on the incident noted the irony with characteristic dryness: “Meta admits its first ‘superintelligence’ was too stupid to survive for three days.”
Sources: Meta admits its first ‘superintelligence’ was too stupid to survive for three days (The Register, Jul 13, 2026)