What is the AI Risk-Scoring Proposal?
The Parliamentary Standing Committee on Finance recently flagged a significant rise in delayed income tax refunds. The number of refunds taking more than 90 days to process has surged, reaching nearly 2.7 million cases in the 2025-26 financial year. To
tackle this, the committee has recommended an AI-based risk-scoring system. The idea is to use technology to automatically separate genuine, low-risk refund claims from suspicious or high-risk ones. Low-risk taxpayers would see their refunds processed much faster, while high-risk claims would be flagged for detailed human scrutiny. This aims to make tax administration more targeted and less intrusive for the vast majority of honest taxpayers.
Decoding 'Risk-Scoring' for Taxpayers
AI-based risk-scoring isn't a new concept for the tax department, which has been increasingly using data analytics to detect discrepancies and fraud. This new system would formalise and expand its use for refunds. The AI would analyse a vast pool of data linked to a taxpayer's PAN, including their filing history, past claims, data from their Annual Information Statement (AIS), GST records, and other financial transactions. Based on this analysis, the system assigns a risk score. A taxpayer with a clean record and consistent data would get a low score, expediting their refund. Conversely, any mismatches, unusual deduction claims, or patterns associated with known fraud would result in a high score, triggering a delay for further verification.
The Lingering Problem of Old Tax Demands
The push for AI comes as the Income Tax Department grapples with two major issues that slow down refunds: a surge in fraudulent claims and a massive backlog of old, outstanding tax demands. The department has noted a rise in bogus deductions, which necessitates deeper verification. However, a more persistent issue for many taxpayers is the adjustment of current refunds against old, often disputed, tax demands. These demands can date back over a decade and often resurface during the digitisation of records, sometimes with accumulated interest exceeding the original amount. Taxpayers complain that these demands appear on the portal without a clear history or proof of service, leaving them scrambling to find old documents to dispute claims they believed were resolved. This process of adjusting refunds against old dues is a key reason for delays, and it's a problem the AI proposal doesn't directly solve.
A Tech Solution for an Administrative Backlog?
While the AI system promises to streamline the processing of new, genuine refunds, critics point out that it operates within a system still burdened by legacy issues. The core problem highlighted by the parliamentary panel's own findings is not just identifying fraud, but also managing the staggering amount of locked-up tax arrears—a significant portion of which is deemed uncollectible. The proposal to use AI to fast-track low-risk refunds is a step towards efficiency. However, it sidesteps the more difficult administrative task of cleaning up the books and resolving decade-old disputes that continue to hold up refunds for otherwise compliant taxpayers. The AI can flag a return for adjustment against an old demand, but it cannot determine if that old demand is valid in the first place. That remains a human, and often lengthy, process of rectification and appeal.
What This Means for Your Future Tax Filings
If this AI-driven system is implemented, the majority of taxpayers with straightforward filings may experience faster refunds. However, the increased reliance on data analytics means that accuracy and consistency in your financial reporting are more critical than ever. Any mismatch between your ITR and the data available with the tax department through AIS and other sources could flag your return, even for small amounts. Furthermore, the issue of old demands remains. Taxpayers should regularly check the income tax portal for any outstanding demands from previous years and address them immediately. Disputing an incorrect demand or paying a valid one proactively is the only way to ensure that a future refund isn't automatically adjusted, regardless of how quickly an AI system processes it.
