Only a few days after launch, Meta withdrew a Muse Image feature that allowed public Instagram accounts to be used as references for AI-generated images. A user could mention a public account in a prompt, and the tool could draw on content from that profile. Meta subsequently updated its official announcement and confirmed that this option was no longer available.
The rapid response is important for brands, agencies and creators. The problem is not that AI-generated images are prohibited in themselves. The problem is the distance between “public content” and “consent to use a person's face, photograph or style in new material.” An option provided by a platform may make a process technically easier, but it does not automatically resolve the rights, agreements or reputational risks surrounding it.
The news also needs to be read accurately: Meta did not withdraw Muse Image as a whole. It withdrew the specific feature through which public Instagram accounts could be mentioned as generation references. The distinction matters because it shows that AI tools can remain available while particular modes of use are stopped when user reaction or the risk of abuse becomes too great.
What happened to the public-account reference feature
Muse Image is the image-generation model Meta launched in July 2026. In its initial form, one option allowed users to mention public Instagram accounts and use them as reference points in images created with Meta AI. The press quickly reported that the mechanism could include public content without separate agreement, while users had to opt out through a setting.
After criticism emerged, Meta said the feature had fallen short of people's expectations and removed it. Reuters and other publications reported that the decision came only days after launch amid concerns about privacy and the use of people's likenesses. The response does not, by itself, establish a universal legal rule, but it provides a very clear example of an AI product that was technically possible yet not sufficiently accepted as a platform experience.
Why “public” does not mean “available for any advertisement”
A photograph published on a public profile can be seen and shared within the platform's limits. It does not automatically follow that a brand can create a new image featuring the person in the photograph, use it in an advertisement, associate the person with a product or reproduce their appearance in AI-generated material.
Several types of rights and obligations may arise simultaneously in a commercial campaign: rights in the photograph, image and likeness rights, the consent of the recognizable person, a creator's contractual rights, advertising rules and the platform's specific terms. Their application differs by country, context and the way the material is used. This article therefore does not replace legal advice for a specific campaign. The operational rule remains simple: a public profile should not be treated as a library of faces for advertisements.
For a company, the risk is not only legal. An image produced without agreement can damage the relationship with a customer, employee, influencer or community. Even if the material is subsequently withdrawn, screenshots, reactions and the brand association may remain.
What a team should check before generating AI images
- The image source. Is it an in-house photograph, licensed material, content received from the client or an image taken from a public profile? The origin must be clear before it is used in a prompt.
- The recognizable person. Is there explicit agreement for commercial use and, where applicable, for AI processing or transformation? Agreement for a photo shoot does not automatically cover every future use.
- The precise purpose. Will the material be used internally, in a presentation, organically on social media or in a paid advertisement? The wider the distribution, the stricter the review needs to be.
- The transformation. Does AI change only the background, or does it create a new representation that may suggest the person endorses a product, promise or message they did not approve?
- Traceability. Retain the source, the agreement, the approved version and the person who authorized publication. A short record is far more useful than someone's memory of a chat conversation.
- Labelling and transparency. Check whether the platform, market or client requires AI-generated or modified content to be labelled, and do not conceal a transformation that could mislead the audience.
Four situations in which brands most often make mistakes
Employee photographs
A company may have permission for a profile photograph or team page. This does not necessarily amount to agreement to use AI to generate new portraits, advertising scenes or videos in which the employee appears to say something they never said. The agreement needs to cover the new use meaningfully, not only the initial presence on the website.
Customer content and testimonials
A customer who tags a brand does not automatically grant a licence for their image to be transformed into advertising material. A clear request for permission, retention of the response and an explanation of how the material will be used are healthier steps than assuming that a public post can be reused freely.
Creators and influencers
A contract may cover specific deliverables: one Reel, several photographs, a period and certain channels. AI reuse can extend beyond those exact limits, especially if the result produces new variations or changes the context in which the person appears. The contract must be reviewed before any experiment.
Images found online
An image from a search engine or public account does not become safe simply because it can be downloaded. For commercial campaigns, in-house sources, libraries with appropriate licences and documented agreements are much easier to manage.
What Meta's withdrawal shows about responsible AI product design
For products that affect people's likenesses, the choice between opt-in and opt-out is decisive. Opt-in means that the person actively chooses to permit a use. Opt-out means that the use is possible until the person discovers the setting and disables it. For a feature that can affect someone's identity or likeness, the difference is not merely one of interface; it is a decision about trust.
Brands using AI have the same responsibility. A tool may have a rapid-generation button, but a sound process also includes an imaginary stop button: who reviews the material, what happens if the person does not give their agreement, and how can an incorrectly published creation be removed quickly?
This is one reason why custom applications with AI integration need to be designed with access rules, permitted sources, logs and human takeover. AI does not become more responsible because it is installed in a new product; responsibility comes from the rules established around it.
What a company can do now
The first step is a short audit of how the team uses images. Identify which materials are owned by the company, which come from customers or partners, which agreements exist and which AI tools have been used for editing or generation. Then establish a simple internal rule: do not use people's faces or photographs in prompts without a clear, documented basis for the intended use.
The second step is to select safer sources: photographs produced for the brand, images with appropriate licences, abstract visual elements, properly photographed products or characters created from scratch without reference to a real person. These provide greater control and reduce the risk that a campaign will appear opportunistic.
The third step is to incorporate review into the social media promotion workflow. An image should not be approved merely because it looks good or was produced quickly. It needs to suit the message, remain coherent with the brand and be usable without creating a debt of trust to the people represented.
Frequently asked questions
Did Meta withdraw the entire Muse Image tool?
No. Meta withdrew the feature through which public Instagram accounts could be mentioned as generation references. Muse Image, the image-generation model, remains announced as a separate product.
If a photograph is public, can I use it in an AI-generated advertisement?
Do not begin with that assumption. Public visibility is not a general commercial licence, and using a recognizable person may require distinct agreements. For a specific use with commercial implications, verify the rights and seek legal advice when there is doubt.
Is it enough if the platform technically allows the use?
No. A product permission or platform setting does not replace an analysis of image and likeness rights, contractual agreements and the way in which the public may interpret the material.
What is the safer option for campaigns?
Use in-house, licensed or entirely original material, retain the agreements and review every AI result before publication. In campaigns involving real people, explicit and documented agreement is the best basis for work.
Sources
- Meta Newsroom – the Muse Image announcement and the 10 July update on withdrawing the option to reference public Instagram accounts.
- TechCrunch – context on the rapid withdrawal following privacy criticism.
- Associated Press – context on the original mechanism and public response.
