By Rhydhi Gupta, Senior Analyst, The Quantum Hub
Last week, Meta launched a feature called Muse Image on Instagram. It let anyone generate AI images of anyone else with a public account, your photos, your face, simply by tagging you. The public response was swift and substantive. Users and privacy advocates objected on three grounds: that the feature reached into people’s photographs by default; that posting content publicly does not amount to consent for AI generation from one’s images; and that such a tool carried an evident risk of enabling deepfake abuse. Within three days, Meta withdrew the feature.
The episode demonstrates that there is considerable uneasiness around the risk of likeness being misused. The potential harms are already apparent and while celebrities have been at the centre of the debate for likeness/personality rights, everyone is at risk. For instance, beginning in 2020, a man in Assam used AI tools to build an online persona from photographs of a former partner, a woman who did not know the account existed. Her face was placed on bodies that were not hers; the content was placed behind paywalls. The account earned approximately ₹10 lakh in subscriptions in the seven months preceding his arrest, and it had crossed 14 lakh followers. As Decode’s subsequent investigation documented, by then her real name had been indexed on pornographic websites, and she withdrew from public life.
This woman is not alone. Official data submitted by India’s Ministry of Home Affairs to Parliament show that complaints recorded on the National Cyber Crime Reporting Portal under sexual-content categories relating to women and children increased from 48,335 in 2024 to 76,657 in 2025. Almost 98% of deepfake pornography globally is directed at women, often created using easily accessible face-swapping applications and automated bot networks. Victims range from school-aged girls to working professionals.
Indian courts have granted personality-rights orders protecting the names, images, and voices of film personalities from AI misuse.
YouTube is building likeness-detection tools for its creators. These developments are welcome, but they extend protection primarily to public figures. It is worth thinking deeply, whether likeness should only be protected for people with a certain privilege. In contrast, Denmark offers a broader model: it is advancing legislation to give every citizen copyright-like rights over their own face, body, and voice, and is advocating the approach at the EU level.
India’s regulatory response has already been strict by international standards. The IT Amendment Rules notified this February require platforms to label synthetic media, to verify user declarations regarding AI-generated content, and to remove non-consensual nudity and morphed imagery within two hours of a complaint, among the shortest timelines anywhere in the world. The framework also imposes proactive duties: platforms must deploy automated tools against certain categories of unlawful content, and the
Non consensual intimate imagery (NCII) SOP proposed by Meity mandates hash-matching and crawler-based detection to stop removed intimate imagery from resurfacing. But these obligations are triggered by a complaint. A removal duty is of limited value without a mechanism for discovery; like a fire brigade without a smoke alarm.
This mechanism can be built. I propose a Likeness ID for everyone: an opt-in system through which any individual, not only a public figure, can register their likeness and receive an alert if it appears in suspected morphed or AI-generated content on participating platforms. What follows the alert is entirely the individual’s choice: she may request a takedown, file a report with preserved evidence to the platform, hand the matter to a trusted contact or support organisation, or to law enforcement if required or do nothing at all.
The system’s purpose is agency, restoring to each person the ability to know how their likeness is being used and to decide what to do about it. The music industry has operated similar content-recognition systems at global scale for almost two decades: rights-holders deposit reference fingerprints, uploads are scanned against them, and enforcement follows automatically. That infrastructure was built because recorded music generates revenue that justifies it. No comparable infrastructure has been built for individuals’ likenesses until now. However India having demonstrated that interoperable public digital infrastructure can be built at population scale, can perhaps build such a protocol across platforms.
It is also worth mentioning that techno-solutionism is tempting here because of scale. However, if implemented there will be tension between protecting the rights of citizens over their likeness and adjacent surveillance concerns that may emanate from likeness IDs. Ultimately, the success of such a solution would hinge on adequate institutional and legal safeguards to make sure that it does not become a face-search engine, a database open to queries by law enforcement or private parties, or a surveillance instrument.
Enrolment to such likeness id would have to be opt-in; and meta data would have to be continuously erased to ensure that a catch-all database is not created for everyone inadvertently. And while we may not have all the answers currently, it is paramount that innovators, industry and regulators sooner than later address this concern through legal or technical means. The pushback against Muse AI by Meta this month shows that ordinary citizens too care about protecting their likeness. The task before regulators is to empower the citizens to protect it.
-Rhydhi Gupta is a Senior Analyst at The Quantum Hub and former Legislative Assistant (LAMP Fellow).