Meta Muse Image Feature Pulled After Backlash

Meta Muse Image launched in early July 2026, but Meta removed an Instagram-linked feature on July 10 after critics raised concerns that public photos were included by default. A Reuters analysis also found its AI image detector failed to identify most cropped AI-generated images, raising questions about content safety and user trust.

Muse Image Rollout and Removal

Meta Platforms introduced Muse Image, its first image-generation model from Meta Superintelligence Labs, as part of an expansion of generative AI tools across its apps. Integrated into the Meta AI chatbot, Muse Image can interpret complex text prompts, accept user photos as inputs, and let users edit generated images with sketches or annotations. Basic use is free, while additional creation features are offered through subscription plans.

The product powers more than 30 new AI effects for Instagram Stories and enables image generation in direct chats with Meta AI on WhatsApp in select countries, with plans to extend the functionality to Facebook and Messenger. These integrations are part of Meta’s broader effort to embed generative features across its ecosystem.

As part of the rollout, Meta introduced an Instagram-linked Muse Image feature that allowed users to generate and modify images using publicly posted Instagram photos. Many public accounts were opted in by default. Private accounts and users under 18 were automatically excluded, and adult users with public accounts could opt out by toggling off settings for Posts and Reels. Meta said it would also act against content violating its Community Standards.

Hollywood talent agencies, managers, and unions including Creative Artists Agency and SAG-AFTRA criticized the design for enabling nonconsensual manipulation of likenesses. SAG-AFTRA called for clear opt-in protections and warned members about risks from unauthorized digital replicas.

Following the backlash, Meta discontinued the Instagram feature that pulled photos from public accounts, acknowledging it “missed the mark.” The company said the capability is no longer available and that it will listen to user feedback while maintaining controls and safety guardrails. SAG-AFTRA welcomed the removal but continued to highlight broader risks posed by AI replicas of likenesses.

AI Detection Limits Exposed

Meta previewed an AI image detection tool using a Content Seal watermark designed to mark AI-generated images. A Reuters analysis of 40 images generated with Muse Image found the detector flagged all original full-size outputs but failed to identify 22 after they were cropped. This represented a 55.0% miss rate on cropped images, reducing detection to 45.0%.

Experts and the analysis highlighted the limits of watermark-based detection: simple edits such as cropping can partially remove or distort the watermark, allowing routine alterations to defeat the detector. This gap between rapid advances in image generation and the robustness of safety tools poses a central technical challenge.

The episode underscores unresolved issues around consent, likeness rights, and the reliability of AI safety systems as Meta pushes to embed generative-image capabilities across its services and compete with other large AI providers. Industry observers say the company will need to strengthen detection methods and rethink consent mechanisms before broader rollouts.