Meta pulls Muse Image @-mention feature after 3-day backlash

Meta shipped its first image model from Meta Superintelligence Labs on July 7, 2026, then disabled one of its headline features 72 hours later. On the evening of Friday, July 10, the company turned off the capability that let users generate AI images by @-mentioning public Instagram accounts and pulling those accounts’ public photos in as visual references.

The reason it matters is not the feature itself but the default behind it: every public profile except users under 18 was opted in automatically, with no notification when their photos were referenced. That is a consent-default decision, and it collapsed within days. For anyone buying or building generative AI tools, the speed of the reversal is the story.

What’s actually new

Muse Image launched inside the Meta AI app, as 30-plus AI effects in Instagram Stories, and in WhatsApp, with Facebook, Messenger, and Advantage+ ad creative slated to follow. The @-mention referencing was one entry point among several. According to TechCrunch’s reporting, users were not told when their images were used as a reference, and the opt-out was described in coverage as a buried settings toggle. That characterization comes from critics and press, not from Meta’s own description.

The under-18 exclusion and automatic opt-in for public accounts were confirmed by local reporting. Backlash built within roughly 48 hours. Users circulated opt-out tutorials, and Hollywood moved fast: on July 8, CAA stated that “no one’s name, image, likeness, voice, or creative work should be used by any third party without clear consent,” and SAG-AFTRA called the launch “an utter miscalculation of public sentiment.” Puck’s Dylan Byers first reported the removal.

Meta’s own statement, appended as a July 10 update to its launch post, is narrow: “We’ve heard the feedback that this feature missed the mark, so it’s no longer available.” Only the @-mention referencing was disabled. Muse Image itself remains live. This is not, despite some framing, a wholesale withdrawal of a product named Muse.

Statement from Meta:

Earlier this week, we announced that one way for people to generate images in Meta AI is by @-mentioning public Instagram accounts that they want to reference. Our intent was to provide a useful creative tool and to give people control over whether their…

— Dylan Byers (@DylanByers) July 10, 2026

What this means for teams shipping generative media

The pattern here is the one that matters for product and legal teams. When the default is opt-out on other people’s likenesses, the rollback clock runs in days, not quarters. Meta learned this in three; OpenAI learned a version of it in 2025, when its Sora launch shipped permissive likeness defaults and retreated under IP-holder pressure. Two of the largest AI shippers have now run the same experiment and gotten the same result.

The lesson is a design one, not a legal one. Defaults are the product decision, a point that holds well beyond likeness: as we noted when Claude Sonnet 5 became the default, the setting most users never change is the one that defines the system. Ship opt-out consent on faces and voices, and the talent agencies, not your roadmap, set your timeline. CAA afterward commended the reversal, saying “putting individual rights and consent at the forefront is essential to building responsible technology,” and SAG-AFTRA called it “the responsible thing to do.”

For enterprises weighing Meta’s ad-creative tools, the takeaway is procedural: consent mechanics and notification defaults now belong in the evaluation, next to output quality. The same governance instinct driving enterprise scrutiny of shadow AI applies to the creative supply chain that feeds these models.

What to watch

Whether Meta reintroduces referencing on an opt-in basis, how the promised Facebook, Messenger, and Advantage+ rollouts handle consent, and whether emerging AI governance bodies treat opt-out likeness defaults as a line worth regulating. No adoption numbers for Muse Image have been disclosed.