VD Satheesan had never performed the dance. His face appeared to disagree.
In a video that hurtled across Instagram, the Keralam Chief Minister could be seen leading Kerala State Electricity Board employees through an energetic routine set to Ala Bolelo, Anirudh Ravichander’s song from Jailer.
The timing supplied the satire. Keralam was confronting a power shortage and criticism over electricity restrictions. Social media responded by making the Chief Minister dance with the very people responsible for keeping the lights on.
There was only one complication. The performance had never happened.
Artificial intelligence had placed Satheesan’s recognisable face upon a body performing movements he had never made. The result was sufficiently convincing—and evidently entertaining—to attract millions of views.
The original Instagram reel posted by AI Kavala carried an AI disclaimer and described the work as entertainment. Indian Express Malayalam reported that it crossed five million views within hours; subsequent reports put the number above 10 million.
The Identity Project #
11 Sep 2026 - Vol 05 | Issue 37
Amit Shah leads the mission for a national demographic correction
Read Now The Identity Project Satheesan watched it too. “I saw the video and found that I was dancing better than Mohanlal and my friend, Tamil Nadu Chief Minister Vijay,” he joked at a seminar on music, cinema and law in Ernakulam. “I was stunned and shocked.”
The laughter then gave way to the law.
“Now Artificial Intelligence is creating music, creating dance, creating everything,” Satheesan said. “How can we interpret in law the new technologies’ creations?”
It is an excellent question because the answer begins with another one. Who owns a dance that never took place?
Was Satheesan’s AI dance illegal? #
There is no automatic answer. The video was labelled as AI-generated. It was political satire responding to an electricity crisis. It did not appear to ask viewers for money, falsely secure Satheesan’s endorsement of a product or place fabricated words in his mouth.
Those facts matter. India’s new regulatory framework does not declare every synthetic image, voice or video unlawful merely because it was created by AI. The Information Technology Rules recognise that synthetic media can have legitimate artistic, educational and humorous uses. MeitY’s official guidance says labelled satire and creative synthetic works may be permitted as long as they do not violate another law.
Satheesan’s video consequently sits close to the boundary between technological impersonation and protected comic expression. The clip uses his identity without recording his performance, but it also announces itself as artificial and operates as recognisable satire. A viewer encountering the original post receives information that the dance is not real.
Remove that label, however, and the character of the video begins to change. Place it inside a fabricated news report and it becomes misinformation. Attach a gambling endorsement and it becomes commercial impersonation. Make Satheesan appear to deliver a communal speech and it could threaten public order. Release a false concession video during an election and the joke becomes an instrument of political manipulation.
The underlying technology does not need to change. Only the intention, presentation and moment of release do.
Does India now have a deepfake law? #
India has deepfake regulation. It does not yet have one comprehensive deepfake statute governing every possible injury.
The distinction is important.
The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, notified on February 10 and brought into force on February 20, introduced India’s first statutory definition of “synthetically generated information”, or SGI.
The definition covers artificially or algorithmically created or altered audio, visual or audio-visual material that appears authentic and depicts a person or event in a way likely to be perceived as indistinguishable from reality.
That includes deepfake videos, AI-generated photographs and cloned voices.
Intermediaries that enable the creation or distribution of such material must deploy reasonable technical measures against unlawful SGI. Permitted synthetic content must carry a prominent visual or audio disclosure and, where technically feasible, embedded metadata or another provenance mechanism identifying its synthetic origin.
Significant social-media intermediaries must ask users whether uploaded material is synthetically generated. They must also use technical systems to test the accuracy of that declaration and display a clear label when the content is found to be synthetic.
These are substantial obligations.
They still do not create a universal personal right under which every Indian can easily say: this is my face, my voice or my movement; you cannot synthetically reproduce it without my consent.
Instead, the victim must identify what additional wrong has occurred. Was the deepfake defamatory? Was it obscene? Did it facilitate fraud? Did it falsely suggest an endorsement? Did it violate privacy, copyright or personality rights? Did it incite an offence or threaten public order?
India has several legal doors. A victim’s immediate problem is knowing which one to open.
Can the three-hour takedown rule stop a viral deepfake? #
It can stop continued circulation at a particular location. It cannot make millions of earlier views disappear.
The 2026 amendments reduced an important intermediary takedown deadline from 36 hours to three hours. But the clock does not begin whenever a person privately complains that a video is false.
It is activated when the intermediary acquires “actual knowledge” through an order from a court of competent jurisdiction or a reasoned written intimation issued by an appropriately authorised government officer.
That is an essential legal safeguard. It prevents every accusation or political objection from automatically becoming a censorship order.
It also creates a practical race.
A synthetic video can be created in minutes, uploaded in seconds, downloaded by viewers, recorded from another screen and reposted across multiple platforms. Meanwhile, the affected person may need to discover the content, preserve evidence, find its original source, determine the applicable offence and secure an order capable of activating the fastest removal obligation.
The Satheesan reel demonstrated the difference between platform time and legal time. The platform measures circulation in swipes. The law measures intervention through procedure.
By the time the three-hour clock starts, the deepfake may already exist in thousands of phones, WhatsApp groups and derivative posts.
Takedown remains necessary. It is no longer synonymous with erasure.
Does a label solve the problem? #
A prominent AI label helps a reasonable viewer understand that a video is synthetic. It does not guarantee that the label will survive the journey.
A user can download a labelled video, crop its border, cover the disclosure, remove the opening audio warning or record the screen. Embedded provenance data can assist platforms and investigators, but metadata may be stripped when content is compressed, edited or moved between services.
The 2026 rules attempt to address this by requiring covered intermediaries not to enable the suppression or removal of synthetic-content labels and identifiers.
The larger vulnerability lies beyond the original platform.
A labelled parody may be reposted without its context. A comic video can become “breaking news” after someone adds a misleading caption. A disclaimer visible on Instagram may disappear when a cropped version reaches a private messaging group.
This creates what might be called the orphaned-deepfake problem. The original has a parent, a label and an explanation. The viral copy has none.
Does Satheesan own his face and movements? #
Indian courts have increasingly said that individuals—particularly celebrities and other widely recognised figures—possess enforceable rights over identifiable elements of their personalities.
In 2023, the Delhi High Court restrained defendants from using Anil Kapoor’s name, likeness, image, voice and other elements of his persona through artificial intelligence, machine learning, deepfakes and face-morphing tools for commercial exploitation.
The court connected unauthorised imitation not only with economic injury but also with privacy, dignity and the effect upon the individual’s family.
Similar orders have protected Amitabh Bachchan and Jackie Shroff. Subsequent cases have expanded judicial protection against AI-generated images, false endorsements, impersonating accounts and the unauthorised commercial use of a person’s name, voice or likeness.
The difficulty is that Indian personality rights have developed largely through judgments drawing upon privacy, passing off, trademark principles, publicity rights and the constitutional protection of dignity.
There is no standalone national personality-rights statute defining precisely what every person owns, how long that right lasts, whether it passes to heirs, what exceptions protect parody and journalism, and when non-commercial synthetic imitation becomes actionable.
That makes the remedy powerful but uneven.
A film star or chief minister can approach a High Court and seek an injunction. An unknown teacher whose voice is cloned for a loan scam, or a college student placed inside a humiliating fabricated video, may lack the resources, knowledge or time to do the same.
AI can clone the famous and the anonymous with equal efficiency. The law does not yet protect them with equal accessibility.
Is copyright the answer? #
Only partly. Copyright protects original works such as films, songs, sound recordings, scripts, choreography and artistic creations. The Copyright Act also grants specified rights to performers over their performances.
But a person’s face is not ordinarily a copyrighted work owned by that person. Neither is the natural sound of a voice. A celebrity’s style, mannerisms or screen presence do not fit comfortably within traditional copyright categories.
The Satheesan video makes the fragmentation visible.
The song has copyright owners. The source dance footage may have separate rights. The choreography may involve creative authorship. Satheesan’s face and identity raise personality and privacy questions. The creator of the synthetic video may claim authorship over its new composition. The platform has intermediary obligations.
One piece of content can therefore activate several different rights belonging to several different people.
Satheesan’s call for stronger copyright protection makes sense in the context of cinema and music, but copyright alone cannot answer the central deepfake question. It protects creative works. The disputed asset here is the human self.
What if the deepfake is satire? #
Any future law must preserve room for parody, criticism, art and political dissent.
Politicians cannot enjoy a legal right to suppress every meme, caricature or synthetic joke that embarrasses them. Public life necessarily attracts imitation and ridicule. A rule broad enough to eliminate all unauthorised depictions of public figures could become a convenient weapon against satire.
India’s 2026 framework acknowledges that lawful, properly labelled satire can continue.
Yet the word “labelled” cannot perform all the constitutional work.
The law must evaluate whether a reasonable person would consider the work authentic, whether the creator intended deception, whether it caused measurable harm, whether it was used commercially, whether it falsely implied endorsement and whether the context made its artificial nature evident.
A visibly absurd AI video of a politician dancing with electricity workers is different from a photorealistic video of the same politician ordering police action. Both are synthetic. Only one may be understood immediately as a joke. The regulatory challenge is to punish deception without placing humour under official licence.
Why are politicians particularly vulnerable? #
A celebrity deepfake can steal money or reputation. A political deepfake can alter public behaviour before anyone establishes that it is false.
An invented speech released on the morning of polling could fabricate an alliance, inflame religious tensions, suppress turnout or make a candidate appear to withdraw. Even if the video is disproved two hours later, some voters will have already seen it and acted upon it.
The harm is time-sensitive and frequently irreversible.
Satheesan’s video became popular because people knew it was a joke about the power shortage. Its comic success depended upon a shared understanding of reality: the Chief Minister had not suddenly begun dancing with KSEB employees.
Political deepfakes become dangerous when that shared understanding collapses.
Once viewers can no longer distinguish documentation from fabrication, authentic evidence also becomes easier to dismiss. A politician confronted with a genuine recording can call it AI-generated. Researchers describe this broader problem as the “liar’s dividend”: the existence of convincing fakes gives dishonest people a new way to deny inconvenient truths.
Deepfakes do not merely create false evidence. They weaken public confidence in real evidence.
What can India learn from other countries? #
There is no universally accepted model.
The European Union’s AI Act uses transparency as its principal response to deepfakes. Article 50 requires those deploying systems that generate or manipulate deepfake images, audio or video to disclose that the material has been artificially created or altered. It provides a lighter disclosure approach for evidently artistic, satirical, fictional or analogous works so that transparency does not obstruct the enjoyment of the work.
Tennessee’s ELVIS Act took a different route. It updated the American state’s personality-rights regime to protect voice as well as name, photograph and likeness from unauthorised AI exploitation. The law was designed particularly around musicians, whose voices can now be imitated without requiring the imitator to reproduce an existing recording.
The US Copyright Office has argued that existing laws do not provide sufficient and consistent protection against unauthorised digital replicas and recommended a federal right targeted specifically at realistic replicas of an individual’s voice or appearance.
Each approach solves a different piece of the puzzle. Transparency tells audiences that something is artificial. Personality rights give individuals control over the commercial or harmful exploitation of identity. Copyright protects creative expression. Platform rules govern distribution and removal. Criminal laws punish fraud, obscenity and other unlawful ends. No single measure can do all five jobs.
So, are India’s laws keeping up with AI? #
They are moving much faster than they were. They are not yet moving as one system.
India’s 2026 rules represent a serious shift from broad intermediary duties to specific synthetic-media obligations. The definition of SGI, mandatory labelling, provenance requirements, user declarations and three-hour compliance window establish a framework that did not previously exist.
But the framework remains primarily platform-facing and reactive.
It regulates how intermediaries handle synthetic content. It does not comprehensively define an individual’s ownership of identity, create an inexpensive universal remedy or resolve the conflict between consent and satire.
Nor can regulation alone overcome the speed of copying.
The next step need not be an indiscriminate ban on deepfakes. India requires a calibrated digital-replica framework that distinguishes declared parody from fraudulent impersonation, protects every person rather than only those able to secure expensive injunctions, and provides rapid remedies proportionate to the likely harm.
It should clearly cover face, voice and other uniquely identifiable characteristics. It should recognise consent, commercial licensing and contractual use. It must protect journalism, art, parody and public-interest research. For election-related deception, intimate imagery, fraud and incitement, it must provide emergency pathways that operate at the speed of circulation.
Satheesan’s dance was funny because the audience was in on the joke. The next deepfake may not offer that courtesy. The Chief Minister never danced. AI supplied the steps, the song supplied the rhythm and social media supplied millions of spectators.
Now the law must decide when imitation is comedy, when identity becomes property—and who is responsible when a person’s digital double refuses to stop performing.
With inputs from ANI & agencies