What Muse Image Is and How It Works
Meta released a new AI image generation model called Muse Image on July 7, 2026 - developed by Meta Superintelligence Labs. According to the technical announcement from Meta AI, it is the company’s most advanced image model to date. It is now available in the Meta AI app, on meta.ai, in Instagram Stories (initially only in the USA), and on WhatsApp in some countries; integration into Facebook is expected to follow.
Technically, Muse Image operates agent-based according to Meta’s own statements: instead of translating prompts directly into pixels, the model calls upon search and coding tools to accurately represent diagrams, QR codes, or interactive content. A self-correction mechanism, which according to Meta emerged during reinforcement learning training, allows the model to evaluate its own results and make targeted improvements. According to the Meta Newsroom announcement, this includes more than 30 new AI effects for Instagram Stories, text rendering for readable infographics, and a tool for redesigning living spaces with product suggestions from the Facebook Marketplace.
For comparison with competing models, Meta provides its own figures: As of July 5, 2026, Muse Image ranked 2nd in text-to-image and image editing tasks in the Arena evaluation based on human preferences, behind OpenAI’s GPT Image 2. Muse Video, which Meta introduced as a preview simultaneously, ranks 3rd in text-to-video.
The Controversy: Photos Without Prior Consent
Shortly after the launch, Muse Image faced criticism due to its default settings. As The National reports, other users can include public Instagram profiles in Muse Image prompts via @-mentions - without the account holder’s prior consent or even notification. This feature is enabled by default, not disabled.
JB Branch, responsible for AI regulation at the consumer protection organization Public Citizen, sharply criticized according to The National: “Meta has once again chosen the creepiest possible path.” His further point, paraphrased: users often only realize afterward that their face was used without permission as source material for someone else’s AI images.
Malwarebytes also warns of another dimension of the problem: public Instagram photos are already being misused by attackers for identity fraud via deepfakes, independent of Muse Image. In combination with generative AI and automated tools, phishing campaigns can be scaled more easily.
How Users Can Protect Themselves
Affected users can disable the feature according to consistent reports from Malwarebytes and The National - but only with several clicks, and the setting is not particularly prominently placed. The path leads through the Instagram settings: Profile > Menu > “Sharing and reuse,” where the “Posts and Reels” option for reuse by AI functions can be turned off. Since the feature is initially being rolled out only in the USA, the setting is not yet visible to all users there.
According to Malwarebytes, an important limitation applies: an opt-out only affects future image generation. AI images created with one’s own face before disabling the setting remain in circulation and are not deleted by the change.
As an additional protective measure, Meta has embedded an invisible watermark called Content Seal in Muse Image outputs, according to its own technical documentation. It is designed to remain intact even after cropping, compression, or screenshots, allowing a separate detection tool to verify whether an image comes from Muse Image. An extension to video is planned according to Meta.
The Business Strategy: Advertising Budgets Instead of Pure Model Quality
Behind the launch, industry observers see a clear commercial logic. According to a Forbes analysis, Muse Image completes the automation of Meta’s advertising platform Advantage+: targeting, bidding, placement, and now also the creation of advertising material are meant to become fully automatable. Meta put the annual revenue of Advantage+ at around $60 billion, with a return of $4.52 per advertising dollar invested as stated by Meta itself, and more than 4 million advertisers already using the company’s generative AI tools.
According to Forbes, the actual competitive advantage lies not in pure image quality, where Muse Image ranks 2nd behind GPT Image 2, but in access to Instagram image material: “Meta has your mother’s birthday photos.” Public photos can thus be specifically integrated into contextually appropriate advertising materials - a capability that competitors without a comparable social network can hardly replicate. Advertisers will gain access to Muse Image through the Advantage+ creative system in the coming weeks, according to Meta’s statements.
Assessment
Muse Image joins a growing number of commercial AI image models with which Meta, OpenAI, and Google compete for users and, above all, for advertising budgets. What is remarkable about this launch is less the technical performance than the combination of default settings that permit the use of other people’s photos and a business strategy that leverages precisely that data access as a competitive advantage. Whether regulatory pressure - for example from the EU with its stricter consent requirements - will change the default settings going forward remains to be seen.
