Meta has shut down one of the key features of its new AI model, Muse Image, which allowed generating images based on public Instagram profiles. The feature, which sparked a storm of criticism, was disabled just three days after its launch.
On July 7, the company introduced Muse Image as the first image generation model from Meta Superintelligence Labs, integrated into Meta AI. The main "highlight" of the new feature was an option: users could specify a public Instagram account in the prompt, and the neural network used its open photos to create new content. However, by July 10, Meta officially updated the announcement, stating that the feature "did not meet user expectations regarding privacy."
The problem lay in the automatic activation of this option — users did not give explicit consent for the use of their images for training or generation. Actress Hannah Einbinder publicly called for the feature to be disabled, and the SAG-AFTRA union (Screen Actors Guild-American Federation of Television and Radio Artists) described the approach as "a complete miscalculation in assessing public opinion." In its statement, the organization emphasized: "Any mechanism other than explicit and conspicuous consent for such uses of Instagram users' images is unacceptable." After the feature was disabled, SAG-AFTRA expressed support for Meta's decision.
Despite this high-profile incident, the Muse Image model itself as a product remains available in Meta AI. It is now used exclusively for generating and editing images based on text prompts and user-uploaded photos. Additionally, Meta has already integrated Muse Image into more than 30 new AI effects for Instagram Stories and into WhatsApp chats with Meta AI in a limited number of countries.
This case is a vivid example of how rushing the launch of AI products can lead to reputational losses. Meta clearly underestimated the audience's sensitivity to the issue of using personal data, even when it comes to public accounts. In an era of total digital awareness, any automatic option affecting privacy is doomed to fail without explicit user consent.