A Deepfake Judge Ran a Fake Court Hearing to Extort a CEO. India Just Moved to Outlaw It.

A deepfake of the Chief Justice ran a fake court hearing

In 2024, SP Oswal, chairman of Vardhman, one of India's largest textile companies, kept a Skype call open for nearly two days while people posing as federal investigators told him he was caught in a money-laundering case. To seal it, they staged a virtual court hearing. Presiding over that fake courtroom was a real-time deepfake of the sitting Chief Justice of India. By the time it ended, Oswal had transferred roughly 70 million rupees, about 729,000 dollars. Police later recovered around 547,000 of it.

Nobody hacked his computer. He was talked out of his money by a video call he believed was official, and the face on screen was manufactured. That single case has become a stand-in for a national crisis, and on 29 July 2026 the crisis reached the top of India's judiciary.

What happened

India's Supreme Court, sitting as a bench of Chief Justice Surya Kant and Justices Joymalya Bagchi and V. Mohana, urged the government to make "digital arrest" a standalone crime and to pass dedicated legislation on AI-generated deepfakes. The matter reached the highest court because no existing Indian law directly names the offense, even after more than 5.5 billion dollars in losses over six years. Solicitor General Tushar Mehta told the bench that a draft bill covering both digital arrests and deepfakes is being prepared. The court cannot write the law itself. Under Article 142 of the constitution it can order justice in a pending case, but only parliament can define a crime, which is why the ruling took the form of a demand to lawmakers rather than a ban.

The scam it targets is scripted. Criminals pose as officers from a body like the CBI (the Central Bureau of Investigation, India's federal police) on a live WhatsApp or Skype video call, tell the target that they or a relative are implicated in a drug or money-laundering case, then place them under "virtual custody" and order them to stay on camera until they pay to clear it. The escalation is the deepfake. Synthetic video built with GANs (generative adversarial networks, a setup where two AI models train against each other until the fake looks real) now runs in real time on ordinary consumer hardware, so a caller can wear the face of a recognizable official. In the Oswal case the gang went all the way to deepfaking the country's most senior judge.

• Digital arrest fraud has cost Indians about 5.52 billion dollars over six years, according to data cited from the National Human Rights Commission, with roughly 2.34 billion dollars lost in 2025 alone as complaints rose 24 percent to about 2.81 million.
• India's cybercrime coordination centre says it has frozen 1.33 million "mule" accounts (accounts rented or sold to launder stolen funds), deactivated 1.2 million SIM cards, and recovered about 572 million dollars.
• Roughly 46 percent of these operations trace back to scam compounds in Cambodia, Myanmar and Laos, many staffed by trafficked workers. India brought home 6,998 of its citizens from such compounds between 2022 and 2025.

Why it matters

The human cost is concrete. Victims include doctors, bankers, retired officials and business executives, and some, worn down by days of coercion and shame, have taken their own lives. Oswal lost the price of a house and got back barely two thirds. The part to carry into your own security meeting is that the trick which put a Chief Justice's face on a video call is the same one that can put your CFO, an auditor, or a "regulator" onto the next call your finance team joins. A familiar face on screen has stopped being proof of who you are actually speaking to. The wider shift is that synthetic-media fraud has gone industrial and cross-border, run from foreign compounds at a scale no domestic police force can arrest its way out of, while the law that would name the crime is still in draft. India operates the world's largest real-time payments network, so the statute it settles on will be studied well beyond its borders. Yet the same government cut its cybersecurity capital budget by 58 percent this year, to about 82 million dollars, at the exact moment its court is demanding more enforcement.

What to do about it

1. Verify any official or executive video request on a channel you control. Hang up and call the institution back on a number you look up yourself, because a real-time deepfake can borrow a trusted face but cannot answer the phone line you dial independently.
2. Make it a standing rule that no real authority arrests, investigates, or demands a transfer over a video call. Brief your staff and your family on it in plain terms, because the whole scheme collapses the moment the target stops believing the call is genuine.
3. Move inside the first hour if money has already left. Report it fast, in India through the 1930 helpline and the national cybercrime portal, because a hold placed on the receiving account within that "golden hour" is far more likely to claw the funds back before they are layered away.

Bottom line

Labeling laws make honest AI content announce itself, which does nothing to a criminal deepfaking a judge to run a sham trial. India's move to criminalize the act itself is the sharper lever, and if it passes it will be among the first laws anywhere to name AI-generated impersonation as its own offense. The defense you can apply before any bill is passed is older and simpler. Treat a face on a screen as a claim, not a credential, and confirm it on a channel the caller cannot reach.