Social platforms and AI are now able to repurpose public photos and even create video likenesses of people without clear notice, raising fresh privacy and ethical questions about consent, data use, and the emotional impact of simulated interactions with the deceased.
Picture scrolling and spotting an image of yourself at a place you never visited or hearing a voice you never recorded; that’s the risk when AI models pull from public social media content. Meta’s short-lived Muse Image program let its AI reference public Instagram posts to generate visuals tied to usernames, and users were not proactively notified about the use. The program was taken down after a swift public backlash, but the episode exposed a broader problem.
The design of Muse Image automatically included any public profile unless the account owner found and navigated a complex opt-out path. That structure meant most people stayed enrolled simply because they didn’t know to opt out, and the practical burden fell on users to protect their own images. The company even described the feature in its announcement blogs with the line, “Whether you want to design a custom event invitation, mock up a collaborative creative concept, or generate a personalized graphic, tagging a username lets Meta AI use public photos to build a visual that’s ready to post.”
Public ire, celebrity complaints, and union pressure combined to force a fast retraction, and Meta issued its statement: “Our intent was to provide a useful creative tool and to give people control over whether their public content could be referenced in this way,” Meta declared in a statement. “We’ve heard the feedback that this feature missed the mark, so it’s no longer available.” Even so, the company has not made clear what happened to the data and images the tool accessed before it was shut down. That gap only deepens people’s anxiety about how their content is handled.
This situation is not unique to one platform; many services default users into data collection that feeds AI training unless they actively change settings. Companies have been using public photos and posts to refine models for years, and once that material is ingested into training datasets it can be effectively impossible to retrieve or erase. The long history of scraping raises a sobering point: once something becomes part of an AI corpus, the window to stop that initial collection often closes.
That history was underscored by critics during past inquiries when Australian Sen. David Shoebridge put the practice bluntly: “The truth of the matter is that unless you have consciously set those posts to private since 2007, Meta has just decided that you will scrape all of the photos and all of the texts from every public post on Instagram or Facebook since 2007 unless there was a conscious decision to set them on private.” Those are strong words, but they capture how sweeping and longstanding the collection has been, and why many users feel blindsided.
Beyond still images, AI now offers the ability to recreate voices and faces on video, bringing another layer of controversy. Techniques that once appeared in major films to resurrect actors for the screen are now accessible outside Hollywood, and some services are being used to generate talking likenesses of loved ones at funerals. “It’s a double-edged sword, as it deals with human emotions,” Yong Man Ro, an AI expert at the Korea Advanced Institute of Science and Technology, told AP, warning that such tools can deliver both comfort and shock.
Legal and psychological scholars caution about the possible harms of simulated interactions with the dead, stressing that technology can interfere with healthy grief. Choung Wan, emeritus professor at Seoul’s Kyung Hee University Law School, put it plainly: “Psychologically, a healthy mourning involves a process to acknowledge the absence of the deceased and pass through the pains of their losses. But speaking with an AI system simulating a living person could undermine the process of accepting deaths and rather cause a negative effect of leaving bereaved families trapped in a fantasy.” Those concerns highlight the emotional stakes, not just the technical ones.
Supporters of resurrection-style tools argue they offer solace and a chance to say goodbye, while opponents point to manipulation risks, consent problems, and lasting mental-health consequences. The rapid spread of these capabilities shows that the technology itself is moving faster than the guardrails meant to protect identity, consent, and privacy. That mismatch between innovation and regulation is the core issue many people are wrestling with today.
The Muse Image episode showed how quickly a company can backtrack under pressure, but it also exposed how easily public content can be swept into model training without users’ clear consent. Even when features are removed, the underlying technology and the datasets collected remain, and questions about retention, use, and control persist. As AI capabilities continue to expand into personal likeness and memory, those unresolved issues will shape public trust and policy debates for the foreseeable future.
