Meta Platforms released Muse Image on Tuesday, the first image-generation model from its Superintelligence Labs unit, embedding it simultaneously in the Meta AI app, Instagram Stories and WhatsApp direct messages. Within hours the launch had produced two distinct news events: a product story and a privacy backlash.

Muse Image, developed under the internal codename Mango, is the second shipment from the Alexandr Wang-led lab after the Muse Spark large language model unveiled in April. It handles text-to-image generation, prompt-based editing, legible in-image text for infographics and QR codes, and a room-restyling tool that pulls listings from Facebook Marketplace. Per CNBC, internal benchmarks put it behind OpenAI’s GPT Image 2 but ahead of Google’s Nano Banana 2 on editing tasks. SiliconAngle reports the model uses “deliberate reasoning” instead of the best-of-N sampling common in image generators, and can invoke a web search tool or hand off coding work to Muse Spark.

The commercial architecture is already in place. Muse Image is free in select markets, with additional capacity gated behind the Meta One subscription plans introduced in May. Advertiser access through the Advantage+ suite is expected in the coming weeks, and a companion Muse Video model is in development.

Then came the posting cycle. A feature that lets users generate AI edits of other people’s public Instagram photos, opt-out by default, drew immediate criticism, per TechCrunch, which framed the default as consistent with prior data-collection controversies: the $5 billion FTC fine paid in 2019 over Cambridge Analytica, and the 2021 shutdown of Facebook’s facial-recognition system.

The pattern is the point. Meta ships the model, ships the default, absorbs the outrage, keeps the model. It’s a playbook the company has run before, and the muscle memory is visible in the release itself.

Sources