Meta’s newly launched Muse Image generator has sparked a major privacy debate on social media, with some users calling the tool a ‘privacy landmine’. While it offers impressive creative tools, one of its features allows the public content from any Instagram handle to be ‘remixed’ into AI-generated images, without any notification to the account.
A section of Instagram users has also flagged that the feature was on by default. Here’s a lowdown of everything you need to know about it and how to claim control over your digital likeness.
What is Muse Image
Muse Image is Meta’s first in-house image generator, replacing third-party models with technology from Superintelligence Labs.
{{/usCountry}}Muse Image is Meta’s first in-house image generator, replacing third-party models with technology from Superintelligence Labs.
{{/usCountry}}It is ‘agentic’, meaning it uses a ‘brain’ called Muse Spark to reason through prompts and plan layouts before the ‘hands’ (Muse Image) draw the result.
The tool is part of apps such as WhatsApp, Instagram Stories, and Meta AI, requiring no new sign-up or credit card information.
Why users are criticising it
Its most controversial feature is the ability to use public Instagram content of users over 18.
Anyone can simply @ your username in a prompt to generate images featuring your face and likeness, with no notification sent to you. Critics have labelled this a “privacy landmine”.
Creative Artists Agency (CAA) strongly opposes Meta's Muse Image, arguing that the default opt-out policy for AI training improperly burdens creators with protecting their likeness from unauthorized use. Demanding an immediate shift to an opt-in model, a CAA spokesperson stated: "No one's name, image, likeness, voice, or creative work should be used by any third party, including AI models, without clear, documented consent".
Key aspects of India's stance and the ongoing assessment include:
Government Assessment: India’s IT Secretary, S Krishnan, stated that the government will assess the tool under the existing legal framework.
The Ministry of Electronics and Information Technology (MeitY) is examining representations received regarding the tool to determine if its features are in accordance with Indian law.
Consent and Default Settings: Indian experts have criticized the "opt-in-by-default" approach for public accounts.
How to check if it's on and opt out of the feature
But, the tool is not all bad!
While Meta’s Muse Image has raised significant privacy concerns, it also offers several advanced features and practical benefits that set it apart from other AI generators:
Readable Text: Unlike many other AI models that produce garbled "AI-soup," Muse Image is notably good at rendering readable text that is spelled correctly for things like birthday cards, invitations, and signs.
Functional QR Codes: It has the unique ability to generate scannable QR codes built directly into a flyer or image.
Steerable Editing: The tool allows for precise, "steerable" edits where you can circle or scribble on a specific part of a picture to erase a photobomber, change the weather, or recolor an object while the AI remembers your previous conversation.
Photo Restoration: It can be used to sharpen and recolor old, faded, or blurry family photos, making them look new again.
Business Mockups: For small business owners, it provides a free way to create product mockups, menu photos, and social graphics without needing professional design skills or software.
Integrated and Free: It is built directly into apps people already use—WhatsApp, Instagram, and Meta AI—and is currently free for casual, everyday creation without requiring a new signup or credit card.
"Agentic" Reasoning: Because it is built entirely in-house, it uses a reasoning model called Muse Spark to plan layouts and look up facts before the "hands" (Muse Image) actually draw the picture, often leading to better results on the first try.
However, there are still some "buts" regarding its limitations:
Not the Fastest: While it ranks high on independent leaderboards, it is sometimes slower than competitors like OpenAI's GPT-Image.
Vague Prompts: It still requires specific instructions to produce high-quality results; vague prompts like "make something cool" often result in generic images.
Watermark Workarounds: While it includes an invisible "Content Seal" to identify AI images, this can be circumvented if someone takes a screenshot of the image