2 months ago
2
A cardinal taxation argumentation situation facing India has been the debased tax-Gross Domestic Product (GDP) ratio and precocious levels of taxation evasion. On average, during 2001-22, India’s tax-GDP ratio (16.36%) was the lowest among emerging and processing economies. India loses astir 4.3% of taxation revenues owed to taxation evasion annually. At the India AI Impact Summit (February 2026), planetary leaders and tech titans hailed India’s advancement successful utilising the powerfulness of Artificial Intelligence (AI) to lick real-world problems.
One occupation country wherever the exertion of AI is worthy examining is taxation gross mobilisation and taxation governance. In this context, India’s Income Tax Department (ITD)’s Project Insight (PI) inaugural — aimed astatine leveraging AI and information analytics to fortify taxation medication and gross mobilisation — warrants captious examination.
The benefits
Launched successful 2017 and afloat operational successful 2019, the PI aims to promote voluntary taxation compliance, trim high-risk cases of imaginable taxation evasion, marque taxation enforcement fairer and equitable, and trim prejudice successful taxation enforcement.
The PI has 3 components. The Income Tax Transaction Analysis Centre (INTRAC) is the analytical motor that utilises AI and precocious information analytics to process fiscal information from aggregate sources, including banking and fiscal institutions, spot and securities transactions, recognition paper and GST payments, and high-value purchases, to make a 360-degree payer fiscal profile.
Such accusation helps the ITD observe inconsistencies betwixt taxpayers’ declared income and their existent fiscal activities. The Compliance Management Centralized Processing Centre ensures behavioural compliance utilizing accusation from the INTRAC. It nudges taxpayers who person filed incorrect taxation returns to record close returns utilizing the Non-intrusive Usage of Data to Guide and Enable (NUDGE) strategy, which involves sending SMS oregon email reminders to the idiosyncratic wage taxes that lucifer their existent economical activities. Taxpayers tin either taxable a revised taxation instrumentality oregon basal by the archetypal returns they filed. Deploying AI successful taxation medication has respective benefits. First, it tin assistance taxation agencies successful accurately assessing taxpayers’ hazard profiles and identifying taxation evasion. Second, it enables taxation administrators to prioritise taxation evasion cases based connected the size and sophistication of evasion. Third, AI tin automate regular taxation medication tasks, freeing taxation administrators to absorption connected those that necessitate greater quality judgment. Fourth, AI tin heighten payer services by assisting taxpayers with filing close taxation returns, answering queries done astute chatbots, and preventing taxation scams.
Outcome of PI initiative
The PI is opening to amusement the results. After receiving nudges, galore taxpayers utilised the ITD’s updated-return diagnostic to marque voluntary changes to their archetypal taxation returns. Since 2020-21, implicit 1 crore revised returns person been filed, resulting successful an further ₹11,000 crore successful taxes. Out of the 19,501 taxpayers contacted by the ITD arsenic portion of a targeted Foreign Income and Assets NUDGE campaign, 62% of them corrected the accusation primitively reported successful their taxation returns. Also, 30,161 taxation filers declared overseas assets totalling ₹29,208 crore and overseas income of ₹1,089 crore from cryptocurrencies oregon virtual integer assets.
The NUDGE run covering 6.25 lakh taxpayers resulted successful corrections of mendacious claims for income-tax deductions amounting to ₹963 crore for governmental donations, and the outgo of further taxes to the tune of ₹410 crore. The mean clip taken to process a taxation refund has decreased from 93 to 17 days. Recently, utilizing large information investigation and AI tools, the ITD discovered that restaurants crossed India had suppressed income turnover of ₹70,000 crore since 2019-20 utilizing blase methods specified arsenic selective deletion of currency invoices, post-billing modifications, wiping of income data, and manipulation of measure value.
Many precocious countries, specified arsenic Australia, Italy, the United Kingdom and the United States person successfully implemented AI-enabled platforms modelled connected the PI and person generated further revenue.
Concerns and risks
As India transitions to algorithmic taxation governance, respective operational, ethical, and ineligible issues necessitate superior attention.
The archetypal is astir information provenance and quality. AI systems specified arsenic the PI are lone arsenic bully arsenic the information they are trained on. They could place outliers, but cannot ever separate evasion from morganatic complexity. Data connected variable-income professionals utilizing anterior savings, clerical errors, oregon associated household fiscal structures tin trigger mendacious positives. Without accessible quality review, taxpayers indispensable beryllium the legitimacy of flagged patterns.
The 2nd is connected the algorithmic bias. AI models trained connected humanities enforcement information tin unintentionally duplicate existing socio-economic oregon geographic biases, identifying taxation fraud much often successful immoderate payer oregon geographic areas than others, arsenic demonstrated by the Dutch childcare benefits scandal.
The 3rd is connected the contented of explainability and owed process. For an AI-based taxation compliance strategy to beryllium legitimate, taxpayers request to cognize wherefore they were identified, however their accusation is used, however the strategy arrives astatine its decisions, and person a clear, casual mode to situation decisions. There indispensable beryllium a human-in-the-loop valuation for immoderate determination with superior consequences for taxpayers.
The 4th is astir the concerns implicit information privateness and security. Accessing delicate fiscal and idiosyncratic accusation astir taxpayers creates a large onslaught aboveground for imaginable exploitation.
Finally, India lacks an AI ombudsperson to reappraisal contested decisions, a request for algorithmic interaction assessments and nationalist reporting of false-positive and appeal-success rates, and outer audits of risk-scoring models.
Without specified beardown AI governance guardrails outlined above, the PI mightiness crook into a hidden surveillance strategy that compromises accountability and taxation strategy fairness, making taxpayers little trusting and much resistant. India indispensable marque a wide choice: to prosecute a modern taxation quality strategy that is some ethical and effective.
Sthanu R. Nair teaches economics and nationalist argumentation astatine the Indian Institute of Management (IIM) Kozhikode, Kerala. Ranjith Gundaboina works connected Artificial Intelligence applications successful fiscal services and was an Assistant Manager astatine the State Bank of India, managing recognition portfolios and hazard assessment. The views expressed are personal
