AI Analysis of Billing Data Uncovers Major Tax Fraud
Indian authorities uncovered a massive tax fraud scheme by using AI to analyze 60TB of billing data from over 177,000 restaurant IDs. The case demonstrates the power of advanced analytics to identify hidden patterns in large datasets, a capability with direct parallels for insurance fraud investigation units.
- The investigation, initiated by the Income Tax Department's Hyderabad unit, uncovered an estimated ₹70,000 crore (approximately $8.4 billion USD) in suppressed sales since the 2019-20 financial year. - Investigators from a digital forensic lab in Hyderabad used advanced analytics and Generative AI to scrutinize the data, which was sourced from a single billing software provider in Ahmedabad that services about 10% of the restaurant market. - The primary method of fraud involved selectively deleting cash invoices after they were issued to customers or wiping out billing records for entire date ranges, sometimes up to 30 days, before filing tax returns. - Out of a total of ₹2.43 lakh crore in billing data analyzed over six financial years, officials found post-billing deletions amounting to at least ₹13,317 crore. - A sample verification of just 40 restaurants in the states of Andhra Pradesh and Telangana revealed a sales suppression of nearly ₹400 crore. - The top five states where the tax evasion was detected were Tamil Nadu, Karnataka, Telangana, Maharashtra, and Gujarat, with Karnataka showing the highest deletions at around ₹2,000 crore. - The Central Board of Direct Taxes (CBDT) has now expanded the probe nationwide, and authorities suspect the current findings may be the "tip of the iceberg" as numerous other billing platforms exist. - In the insurance sector, AI is similarly used to detect fraud by analyzing vast datasets of claims and policy information in real-time to identify suspicious patterns, anomalies, and inconsistencies that would be missed by human review.
