evidoria

← Back to browse

Good practice

GSTN BIFA & ADVAIT — India’s AI-powered GST fraud-detection platform

India · New Delhi · See the India profile

India’s GSTN uses BIFA and ADVAIT AI analytics tools to detect GST fraud across ~130 million monthly invoices. The GST enforcement apparatus detected ₹7.08 lakh crore (~USD 84 billion) in evasion across 91,370 cases over FY2020–25, tabled by the Minister of Finance in Parliament.

~130 million per month
Invoices processed
~3 billion per month
API calls
₹7.08 lakh crore (~$84 billion)
GST evasion detected (FY2020-21 to FY2024-25)
91,370
Evasion cases (FY2020-21 to FY2024-25)
₹49,384 crore
Evasion detected, FY2020-21
₹2.30 lakh crore
Evasion detected, FY2023-24
₹1.79 lakh crore
Evasion attributable to ITC fraud

Details

Maturity
Established
Promoter
Goods and Services Tax Network (GSTN)
Period
2017–present
Keywords
GST compliance, fake invoicing, ITC fraud, machine learning, big data analytics, e-invoicing

Context

GSTN is India's not-for-profit technology backbone for the nationwide GST system, processing approximately 130 million invoices and 3 billion API calls every month; it has progressively integrated AI/ML tools BIFA (Business Intelligence & Fraud Analytics) and ADVAIT alongside e-way bill analytics.

Objectives

To identify risk patterns, fake invoicing networks, and fraudulent Input Tax Credit (ITC) claims by creating risk profiles for every registered taxpayer, letting DGGI and state enforcement authorities prioritise limited audit resources.

Activities

The e-way bill analytics layer cross-references movement-of-goods data with invoice declarations in real time, based on filing history, ITC mismatches, turnover trends, and sector benchmarks.

Results

Data tabled by the Minister of State for Finance in the Lok Sabha in August 2025 showed GST enforcement — supported by these AI tools — detected ₹7.08 lakh crore (~$84 billion) in evasion across 91,370 cases over FY2020-21 to FY2024-25, rising from ₹49,384 crore in FY2020-21 to ₹2.30 lakh crore in FY2023-24; ₹1.79 lakh crore was attributable to ITC fraud.

Conclusions

The algorithms are not publicly disclosed, no independent bias audit has been published, and the Parliamentary figures reflect total enforcement output rather than outcomes attributable solely to the AI tools.

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful