Meta shipped Muse Image on Tuesday, its first in-house AI image-generation model, embedding it simultaneously inside the Meta AI app, Instagram Stories, and WhatsApp direct messages. Within hours, the ship narrative had a second story stapled to it: a default-on setting that lets any user tag a public Instagram account and remix that person’s photos into AI images, with no notification to the account owner.
Muse Image, code-named Mango internally, is the second release from Meta Superintelligence Labs, following the Muse Spark LLM that succeeded the Llama family in April. Alexandr Wang, Meta’s Chief AI Officer, was hired to lead the lab roughly a year ago, after Meta invested billions rebuilding it and recruited researchers with compensation packages exceeding $200 million. The product cadence is now the point Wang is being paid to prove.
Consumers get free access via the Meta AI app, WhatsApp, and Instagram Stories, with paid tiers routed through the monthly subscription plans Meta introduced in May. Advertisers get Muse Image inside Advantage+ in coming weeks. A Muse Video model is, per TechCrunch, “already in development.” Bloomberg reports Meta also plans to sell Muse Image to outside developers as a cloud offering, of a piece with last week’s reporting that the company’s data-center buildout could extend to selling raw compute.
The technical pitch, per SiliconAngle, is that Muse Image uses “deliberate reasoning” rather than the best-of-N sampling most image generators rely on, supports test-time compute scaling, and can call a web-search tool when a prompt lacks detail. Outputs can hand off to Muse Spark to become websites or vibe-coded games. Meta’s own benchmarks place the model behind OpenAI’s GPT Image 2 but ahead of Google’s Nano Banana 2 on editing tasks. Nano Banana became a consumer hit last fall, which is the comparison Meta actually cares about.
Then there’s the tagging feature, first flagged by The Verge. Meta’s policy language reads that “people may be able to create content with your Instagram content using AI features at Meta” and that “you will not be notified about content created using AI features at Meta.” The opt-out requires digging into settings. One X user, quoted by TechCrunch, called it “a privacy landmine waiting to detonate.”
The framing is doing work Meta can’t fully control. In 2019 the company paid a then-record $5 billion FTC fine over Cambridge Analytica. In 2021 it shut down Facebook’s facial-recognition system amid biometric-data lawsuits. Meta says Muse Image outputs carry an invisible watermark and that the system blocks CSAM and other terms-of-service violations.
The pattern is the story. Meta keeps shipping consent architectures that require users to find the switch, and keeps discovering the switch is where the backlash lives.
Sources
- https://www.bloomberg.com/news/articles/2026-07-07/meta-debuts-new-ai-image-generation-model-inside-chatbot-instagram
- https://www.cnbc.com/2026/07/07/meta-ai-muse-image.html
- https://techcrunch.com/2026/07/07/meta-rolls-out-muse-a-new-ai-image-generator/
- https://thenextweb.com/news/meta-muse-image-ai-generation-instagram-whatsapp
- https://siliconangle.com/2026/07/07/meta-launches-image-generation-model-coding-search-capabilities/