The writer elaborates on how AI is fighting Goa’s cyber fraud epidemic
From a fake video call to a police lab in Ribandar, the machines are now on both sides
11:40 on a Thursday at a manufacturing unit off Verna. The accounts executive joins a video call. The managing director is on screen, two colleagues beside him, asking for an urgent transfer against a confidential order. The voice is right. The face is right. Money moves in four tranches before lunch. Nobody on that call was real.
That is not a Goan case yet. It happened to the engineering firm Arup, whose Hong Kong finance staffer wired HK$200 million, roughly US$25 million, across 15 transfers in January 2024 after a video conference in which every participant except him was AI-generated. The firm confirmed it that May. The technique has since become cheap.
Goan business has always run on knowing the voice at the other end. That assumption is now a liability. My wager: the first deepfake-led transfer fraud reported in Goa lands on a mid-sized firm with a busy accounts desk, not a bank, and it lands inside two years
What Goa is already losing
Goa does not need deepfakes to have a problem. Cybercrime cases registered in the state went from 82 in 2023 to 85 in 2024, then jumped to 131 in 2025. More than 1,500 cyber fraud complaints were filed in the first four months of 2026. The Union Home Ministry told the Lok Sabha last month that Goa has logged 446 e-Zero FIRs for cyber financial fraud since that system went live here this year.
The individual numbers are worse. A Quepem resident lost `61.84 lakh to a fake ‘digital arrest’ run by callers impersonating a Mumbai Crime Branch officer. A Panjim resident lost `60 lakh to the same script, with accomplices playing CBI officers and a magistrate. The Chief Minister puts Goan losses to ‘double your money’ investment schemes at `500 crore to `1,000 crore, with several of the operators working out of Dubai and London.
Nationally the shape is identical. Ministry of Home Affairs figures put 2025 losses at `22,495 crore across 28.15 lakh reported cases, with investment fraud accounting for 76% of the money lost and digital arrests another 9%.
The defence is already an AI stack
What most Goan business owners have not registered: the machinery now intercepting this money is machine learning, and it works.
The Department of Telecommunications runs a Financial Fraud Risk Indicator that scores mobile numbers as medium, high or very high risk, drawing on the national cybercrime portal, the Chakshu platform and bank intelligence. Since rollout on 22 May 2025, it prevented roughly `660 crore of losses in six months, with over 1,000 banks, payment apps and payment system operators onboarded. RBI directed all scheduled commercial, small finance, payments and co-operative banks to integrate it on 30 June 2025.
The Reserve Bank Innovation Hub’s MuleHunter. AI attacks the receiving end. Built after its developers mapped 19 distinct patterns of mule account behaviour with banks, it flags around 20,000 mule accounts a month; 23 banks had it live by December 2025. Separately, the I4C suspect registry has pooled over 27 lakh mule accounts and declined transactions worth `9,518 crore.
Goa is not a spectator. The state police operate an AI-ML Lab at the Crime Branch in Ribandar, inaugurated in February 2025 with a Goa-based startup as technology partner, and have built DeepTrace, which the force describes as the first AI-powered investigative platform developed by any state police force in India. On 2 August, around 40 Goa Police teams raided over 40 locations across 11 states, arrested seven people, and verified 313 mule bank accounts linked to over `2.5 crore of fraud registered in the State.
What you do on Monday
Not buy software. Three things, two of them free. Write a callback rule and make it absolute: no payment instruction from any voice, face, email or WhatsApp message gets executed until someone rings back on a number already sitting in your records. The Arup fraud unravelled the instant an employee called head office. That control costs nothing and would have stopped the most expensive deepfake in corporate history.
Report suspicious calls and messages on Sanchar Saathi. Those citizen reports feed the risk scores already declining transactions across the banking system.
And 1930, at once. Freezing works in the first hour, not the first day.
Goan business has always run on knowing the voice at the other end. That assumption is now a liability. My wager: the first deepfake-led transfer fraud reported in Goa lands on a mid-sized firm with a busy accounts desk, not a bank, and it lands inside two years.
The national stack will catch the money afterwards, sometimes. The callback catches it before