Just three days after the high-profile launch of its new image generation model, Muse Image, Meta Corporation was forced to withdraw one of its key features, which sparked a wave of criticism. The feature in question allowed the neural network to use public photos from Instagram to create new images based on a text prompt.
Why did the feature fail?
Initially, Muse Image, unveiled on July 7 as Meta Superintelligence Labs' flagship product, offered a unique capability: adding the name of a public Instagram account to a prompt so the AI could analyze the user's open photos and generate new content based on them. However, by July 10, the company quietly removed mention of this feature from the official announcement, acknowledging that it "did not meet user expectations regarding privacy."
The problem turned out to be deeper than just a technical flaw. The option was enabled by default, causing outrage not only among regular users but also among professionals. Actress Hannah Einbinder publicly called for the immediate deactivation of this feature, and the influential union SAG-AFTRA issued a sharp statement. In their view, any mechanism other than explicit and informed consent for the use of images constitutes a "complete miscalculation" and ignores the obvious risks associated with digital identity and deepfakes.
What remains of Muse Image?
Despite the scandal, the Muse Image model itself has not disappeared. It continues to operate as part of Meta AI, allowing users to generate and edit images based on text descriptions and uploaded photos. Moreover, the technology has already been integrated into over 30 new AI effects for Instagram Stories and is used in WhatsApp chats with Meta AI in several countries. However, the lesson taught by the community is clear: even the most advanced algorithms should not operate at the expense of privacy without the user's explicit consent.
Cryptalist Analysis: This incident is a classic example of how the pursuit of innovation can lead to reputational losses. Meta once again faced the "uncanny valley effect" in AI ethics. By disabling the feature, they acknowledged that user control over data is not an option but a fundamental requirement. For the industry, this is a signal: any product that uses public data for training or generation must be designed from the outset with the principle of "privacy by design," rather than being corrected post-factum under public pressure.