Rs 70,000-crore scam? How Hyderabad I-T dept used AI to flag pan-Indian tax evasion at restaurants

The catalyst? AI tools, which were used to analyse a whopping 60 TB (60,000 GB) worth of billing data from a software used for more than 1 lakh restaurant transactions

food-payment-tax-evasion-scam - 1 Representative image of a person making a card payment | Shutterstock

What began as an in-depth probe into the revenue of biryani restaurants in Hyderabad has snowballed into a larger investigation into a pan-Indian tax evasion scam worth thousands of crores, a report has said.

The catalyst? Artificial intelligence (AI) tools, which were used to analyse a whopping 60 TB (60,000 GB) worth of billing data from a software used across India for more than one lakh restaurant transactions, or about 10 per cent of the restaurant software billing market.

According to a Times of India report, the Hyderabad wing of the Income Tax Department has revealed that the restaurants involved across India had suppressed at least Rs 70,000 crore worth of sales turnover since FY 2019-20.

Across India, restaurants using this software were found to have erased bills worth over Rs 13,000 crore. Of this figure, Rs 5,141 crore in hidden sales was said to have come from Andhra Pradesh and Telangana.

To double-check this second figure, officers visited 40 restaurants on the ground and compared real sales with computer records. Notably, even this revealed nearly Rs 400 crore in missing sales, the report said.

It added that the most instances of tax evasion were detected in Karnataka at Rs 2,000 crore, Telangana with Rs 1,500 crore, and Tamil Nadu at Rs 1,200 crore. Maharashtra and Gujarat were the other states that also saw many cases of tax evasion.

How it was done

The I-T department investigators noted that restaurants typically enter all types of sales—card, UPI, or cash—into the software they analysed to prevent internal manipulation by servers, cashiers, or managers.

One common pattern they flagged was the restaurants allegedly selectively deleting invoices—keeping only a portion of cash entries and deleting the rest to reduce income tax and GST exposure, they said, citing internal documents.

Another pattern they found was the deletion of bills in bulk—wiping bills clean for a chosen date range—including up to 30 days of billing—followed by filing returns adjusted to only a fraction of the actual sales.

Covering billing worth Rs 2.43 lakh crore across six financial years, the investigators used AI tools loaded on high-capacity systems to quickly map GST numbers to restaurants using open-source information and publicly available online material, and then to analyse the monetary data itself.

The probe is now expected to extend to other billing platforms to get a better picture of the extent of tax evasion in the Indian restaurant space.