What Exactly Is the Proposal?
The Parliamentary Standing Committee on Finance has recommended using an Artificial Intelligence (AI)-based risk-scoring system to process income tax refunds. The suggestion comes in response to a significant increase in delayed refunds, which climbed
to nearly 2.7 million cases in the 2025-26 financial year. The panel, chaired by BJP leader Bhartruhari Mahtab, believes this tech-driven approach could separate genuine, low-risk refund claims from suspicious ones, allowing honest taxpayers to get their money back faster. The tax department has stated that many delays are due to essential checks for bogus deductions and exemptions, unvalidated bank accounts, or mismatches between PAN and Aadhaar details. The committee's proposal aims to automate this sorting process, focusing human scrutiny on high-risk cases while clearing the vast majority of claims without delay.
How Does AI Risk-Scoring Work?
Think of it as a highly advanced filter. An AI risk-scoring system would analyze a massive amount of data to assign a risk score to each tax return. This data includes not just the information on your return, but also your financial history from sources like the Annual Information Statement (AIS), which aggregates transactions from banks, stock exchanges, and property registrars. The AI looks for anomalies, inconsistencies, and patterns that might suggest an error or a fraudulent claim. For example, it might flag unusually high deduction claims that don't match your income profile or discrepancies between your declared income and the data available with the tax authorities. A low-risk score would mean the refund is processed automatically and quickly, while a high-risk score would trigger a manual review by a tax officer.
The Promise: Efficiency and Fraud Detection
The government's primary motivation is to tackle two major problems: processing delays and tax evasion. India's tax administration already uses AI in projects like 'Project Insight' and GST analytics to identify non-filers and fraudulent activity. By extending this to refunds, the goal is to make the system more efficient, reducing the working capital stress on individuals and small businesses waiting for their money. Proponents argue that AI can process millions of returns far faster than humans, spotting red flags that might otherwise be missed and curbing the large-scale GST evasion and fraudulent Input Tax Credit (ITC) claims that the committee also highlighted. The system could also nudge taxpayers toward better compliance by providing real-time feedback on potential discrepancies before a return is even submitted.
The Risks: Bias and the 'Black Box' Problem
Despite the potential benefits, using AI in tax administration carries significant risks. The biggest concern is algorithmic bias. If the historical data used to train the AI contains biases, the system could unfairly target certain groups of people—for instance, those from specific regions, income brackets, or professions—leading to disproportionate scrutiny. This raises constitutional questions about the right to equality. Another major issue is the 'black box' problem, where the AI's decision-making process is so complex that it's difficult for even its creators to explain why a specific return was flagged. This lack of transparency can make it nearly impossible for a taxpayer to understand, let alone appeal, a decision. Without robust safeguards, there's a risk of creating a system that is efficient but also opaque and potentially unfair.
The Way Forward: Balancing Act
The parliamentary panel's proposal is not a final decision but a strong recommendation that forces a crucial conversation. Experts agree that human oversight will remain essential; AI should be a tool to assist tax officers, not replace their judgment. As India moves toward a more data-driven tax ecosystem, establishing clear rules for transparency, accountability, and fairness will be paramount. This includes creating formal appeal processes for AI-driven decisions and conducting regular audits to check for bias and accuracy. The proposal makes it clear that the future of tax administration involves AI, but the challenge lies in designing a system that balances the government's need for efficiency with the citizen's right to a fair and transparent process.
