The Borrowed Face and the Shrug

Tulika

We grew up learning that a picture is worth a thousand words. That may no longer be true. Thanks to AI, we have entered an age where society is being taught that believing a picture is the most foolish thing you can do.

That lesson is half right. The other half is where people are now getting hurt. Refusing to believe any picture is as foolish as believing each one of them. The useful question was never “is this real?” It is “where did this come from, and can I check?”

This post is about the people in India who learned that the hard way.

Two words we keep mixing up

AI-generated images, or synthetic media, is the broad umbrella. It covers any picture, audio clip, video or text that a machine learning model has created or substantially altered. A misty Himalayan valley that never existed. A portrait of your cat in the style of a Mughal miniature. A stock photo of a smiling office team, made from one line of prompt. Most of it is harmless. Some of it is lovely.

A deepfake is a narrower thing inside that umbrella. The word joins “deep learning” and “fake”. It means synthetic media that realistically mimics a real, identifiable person. Their face is swapped onto someone else’s body. Their mouth says words they never spoke. Their voice is forged.

The difference is a person. A made-up valley harms no real person. A deepfake borrows a real human being without asking and makes them do something they never did, so the harm lands on a named individual who has to live with it long after the clip has stopped trending.

Indian law now draws a similar line. The IT Rules, amended in February 2026, cover “synthetically generated information”. That means audio, visual or audio-visual content made or altered by a computer “in a manner that appears real and authentic”. Text is not in that definition. The test is whether it looks real.

A black ink cartoon of an Indian living room. On a sofa, a young man in a T-shirt holds up a phone while an older woman in a sari clutches her cheek and a middle-aged man with a moustache jabs at his own phone, all three gasping. At the far right, a man in spectacles sits apart on a chair, looking at his phone and writing in a small notebook. — Three of them have seen it. One of them is asking who made it.

A classroom in Guwahati

In June 2026, two well-known schools in Guwahati faced the same kind of complaint. A male student had allegedly used an AI tool to make explicit images of girls in his school, and of their teachers. The images went round among classmates and onto online platforms. Some, according to reports, were shared for money.

The first school opened an inquiry and suspended the students directly involved. A second school reportedly suspended a student over similar images. Assam Police’s Cyber Crime Branch began an investigation.

None of the girls posed for anything. None of the teachers did either.

This is the part of the story that is about us, not about AI. In the past, a teenager who wanted attention had no means to do this. Today he has a phone, a free app and a group chat that rewards whatever gets a reaction, and the cheapest reaction to get is a picture of someone he knows in a situation she never chose, so the urge to hurt someone now arrives dressed as a joke.

The harm does not stay in the chat. A girl walks into class knowing that some of the boys have seen her, even though they have seen nothing real. A teacher stands in front of a room and wonders which of them has the picture.

A face she trusted

The next case had nothing to do with school.

In February 2025, a 57-year-old retired woman in CV Raman Nagar, Bengaluru, watched a YouTube video. In it, a well-known spiritual teacher appeared to recommend an online trading platform. Start with 250 US dollars, the video said, and your finances will improve. It was a deepfake.

She clicked the link and filled in her details. A man calling from a UK number asked her to download a trading app. He and a colleague then ran “lessons” for her on Zoom. Between February and April, she transferred Rs 3.75 crore. She found out only when she tried to withdraw her profits and could not. She filed a police complaint in September.

Notice what failed. A face she trusted was no longer proof of who was speaking. The scammers did not need to fool her for long, because the deepfake only had to open the door, and after that the Zoom calls, the dashboard showing rising numbers and the friendly voice from abroad did the rest, one small payment at a time.

The same trick runs at national scale. In January 2026, BSE warned investors about a deepfake video of its own managing director handing out stock tips and promising viewers they could become “multi-millionaires within a short time”. He had said no such thing.

The opposite mistake

Here is where the lesson from the opening goes wrong.

The law professors Robert Chesney and Danielle Citron named the problem in 2019. They called it the liar’s dividend. Once everyone knows fakes exist, anyone caught on a real recording can simply say “deepfake”. If people have learnt to disbelieve every picture, they will just shrug. So the people who gain the most from doubting everything are not careful citizens. They are the ones with something to hide.

It happens closer to home than courtrooms. A real video of a cousin driving rashly reaches the family group. The cousin replies with two letters: AI. Half the group believes him, because by now that answer always sounds possible.

So a society that believes everything is easy to fool. A society that believes nothing is easy to fool too. Only the method changes.

Still happening

None of this is history. In May 2026, a clip face-swapping the actor Madhuri Dixit onto awards-show footage spread on X. A Community Note appeared only after it had travelled widely. In October 2026, an explicit deepfake of the actors Janhvi Kapoor and Jr NTR spread online. Kapoor called it “deeply violating, bordering on sexual assault”.

The rules are stricter than they have ever been. Since February 2026, platforms must label synthetic content and act on takedown orders within three hours. Non-consensual intimate images have a two-hour window. Yet MediaNama counted at least 19 unlabelled AI videos shared by political parties during the 2026 state elections.

Law moves at the speed of a gazette. A forward moves at the speed of a thumb. In the long gap between a rule being notified in Delhi and a clip landing on a phone in a school group chat in Guwahati or a family WhatsApp group in Bengaluru, it is the person holding the phone who decides whether the clip spreads or stops.

The form I did not fill

My own brush with this involved no AI at all. It did not need to.

Some time back, I bought a flat in Noida. I had to send a part payment by NEFT. I was a little unsure of the process, so I called my bank for guidance.

Soon after, my phone rang. The number was unknown. The caller said the transfer would go through as soon as I filled in a few details in an online form. Then a Google Form link arrived.

It sounded right. I had just called the bank, and here was someone calling back about exactly what I had asked, at exactly the moment I was waiting for it, which is the whole trick and the same one a deepfake plays: it shows you what you already expected to see.

My suspicion got the better of me and, praise the Lord, good sense prevailed. I called the bank back on their official number and it was then I got to know that this was the work of an organised gang of scammers.

I did not need a detector. I needed to put the phone down and call back on a number I already trusted.

What to do on Tuesday

Five habits. None of them needs software.

Before you forward, ask where it came from. Not who sent it to you. Who made it. If you cannot find the first source, you have not verified anything.

A famous face in an advertisement is not an endorsement. Before you invest a rupee, check whether the platform or adviser is registered with SEBI. Its website lets you search.

When a familiar face or voice asks for money, hang up and call back on a number you already had. Agree on a family word, something only your people know. A cloned voice can copy your son’s accent. It cannot guess the name of the dog you had in 2009.

Talk to your children about the group chat. Not only about what they receive, but about what they forward. If you have been cheated or targeted, report it at cybercrime.gov.in or call 1930, the national cybercrime helpline.

And when someone you dislike is caught on camera and says it is a deepfake, do not believe them just because you want to. Ask for the same evidence you would ask of a stranger.

A picture is still worth a thousand words. We just have to ask who wrote them.

October 13, 2026 | c.tulika2309@gmail.com

Sources

IT Rules amendment on synthetically generated information, notified 10 February 2026, in force 20 February 2026: Ministry of Electronics and Information Technology, Gazette notification, summarised by Hogan Lovells, “India introduces mandatory labelling for AI and 3-hour takedown for illegal content”.

Guwahati schools: Free Press Journal, 12 June 2026. Police investigation and images shared for money: MediaNama, “10 cases that show India’s deepfake rules are not being enforced”, June 2026.

Bengaluru trading fraud: The Tribune, 11 September 2025.

BSE warning: IANS, “BSE warns investors about fake deepfake video misusing CEO’s identity”, 12 January 2026.

Liar’s dividend: Robert Chesney and Danielle Citron, “Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security”, California Law Review, vol. 107, 2019.

Madhuri Dixit clip, election videos and two-hour NCII window: MediaNama, June 2026, as above. Janhvi Kapoor and Jr NTR: MediaNama, October 2026.

Was this worth your time?