The Push for Faster Refunds
For years, Indian taxpayers have voiced frustration over delays in receiving their income tax refunds. A recent report from the Parliamentary Standing Committee on Finance highlighted the scale of the issue, noting a sharp increase in refunds taking longer
than the prescribed 90-day period. In the 2025-26 financial year, nearly 2.7 lakh cases faced such delays. The committee, chaired by MP Bhartruhari Mahtab, identified these bottlenecks as a cause of cash flow stress for individuals and businesses. To solve this, the panel has recommended a technology-driven solution: an Artificial Intelligence (AI) based system to streamline the process. The goal is to separate genuine claims from suspicious ones, allowing the vast majority of honest taxpayers to get their money back quickly while high-risk cases are flagged for deeper scrutiny.
How AI Risk-Scoring Would Work
The proposed system would function like a digital gatekeeper. When a taxpayer files a return and claims a refund, an AI algorithm would analyze their profile. It would look at their filing history, the nature of deductions claimed, and cross-reference information from various financial sources to generate a 'risk score'. A taxpayer with a clean record and standard claims would likely be classified as 'low-risk', and their refund would be processed almost instantly. Conversely, a return with unusual deductions, inconsistencies, or other red flags would be scored as 'high-risk' and automatically routed for manual review by a tax officer. This is part of a broader push by the tax department, including 'Project Insight', to leverage data analytics for better tax administration. The idea is to make the system more efficient and less intrusive for compliant taxpayers.
The Hidden Catch: Old Tax Demands
Here's where the forward-looking AI proposal collides with a messy legacy problem. Under Section 245 of the Income Tax Act, the tax department has the power to adjust any refund due to a taxpayer against any outstanding tax demand from a previous year. This is a powerful recovery tool. The problem is that the system is clogged with lakhs of old, often disputed, tax demands. Some of these demands date back 10 or 15 years and have resurfaced due to the department's efforts to digitize records. Taxpayers often claim they were never properly notified of these demands or that the demands themselves are incorrect and under appeal.
The Fear of Automated Coercion
The primary concern is that the AI risk-scoring system will automatically flag any taxpayer with an outstanding demand—regardless of its legitimacy—as 'high-risk'. This could effectively automate the process of withholding refunds. A taxpayer could have their legitimate refund for the current year blocked by a decade-old demand of a few thousand rupees, which may have ballooned with accumulated interest. While the law requires the department to send an intimation notice before making such an adjustment, the process puts the onus on the taxpayer to respond and fight an old battle, often for a demand they believe is invalid. Critics fear the AI will become a tool of coercion, forcing taxpayers to pay up on disputed legacy demands just to unlock their current refund, turning a system meant for efficiency into one of automated harassment.
Balancing Efficiency with Fairness
The government has made some attempts to address the mountain of old dues. In early 2024, the CBDT announced a scheme to extinguish certain small tax demands up to a cumulative limit of ₹1 lakh per taxpayer for periods up to the 2015-16 assessment year. However, this only addresses a fraction of the total outstanding amount, which was reported to be a staggering ₹47.42 lakh crore as of January 2026, though a large portion is deemed 'uncollectible'. The parliamentary panel's recommendation for AI brings this tension into sharp focus. While technology can undoubtedly speed up processes, its implementation must include robust safeguards. Without a mechanism to distinguish between a confirmed, undisputed tax liability and a contentious demand that is under appeal, the AI system risks penalizing taxpayers for systemic administrative backlogs and unresolved disputes.
