The Income Tax department’s increased use of artificial intelligence (AI) and machine learning (ML) tools in processing tax returns has led to a flurry of automated notices and text messages to taxpayers, indicating some discrepancies in their tax filings.
Under the e-verification scheme, the tax department is issuing notices to employers to verify discrepancies in TDS deducted by them on the salaries of respective employees based on the apparent incorrect HRA declarations by the employees.
Discrepancies are flagged when the PAN of the landlord in the rent receipt, used for claiming deduction in House Rent Allowance (HRA), is either invalid or linked to a person not accounting for the rental income in their return.
It has now become crucial to furnish accurate HRA declarations, and employers should verify these with rental agreements containing complete and authentic credentials of the landlord.
In case of non-availability of such rental agreement, a signed confirmation from the landlord, with PAN, acknowledging the fact of receipt of such rental income, should be asked. In the absence of the same, employees should be advised that HRA benefit may not be given in TDS deduction on their salaries.
The IT department also sends an automated SMS to taxpayers to revise their returns for mismatch between disclosures in returns and information received from the reporting entities, in respect of certain high-value specified financial transactions (SFTs).
SFTs include foreign remittances under liberalised remittance scheme (LRS) captured by tax collected at source (TCS), purchase of immovable property captured by TDS and Annual Information System (AIS), transactions in securities, mutual funds, derivatives and commodities captured by AIS, exports, imports, acquisition of any foreign asset, renewal of bank fixed deposit in excess of ₹50 lakhs, foreign travel, high-value credit card spends.
Taxpayers should respond via the IT department’s compliance portal by selecting an appropriate response out of the four available options — information is correct, information is duplicate/included in other PAN, information is incorrect and the residual category response in ‘others’.
The main objective of this e-campaign is to encourage voluntary compliance by the taxpayers and provide them with an opportunity to rectify defects, if any, in their returns by filing a revised/belated return, before the deadline of 31 December, for the returns filed for assessment year 2023-24.
However, the over-dependence on AI and ML tools by the IT department leads to automated red flags, without involving any human vetting. There are several instances, where the same financial transaction gets repeated in the AIS, and as such multiplying manifolds in value, and thereby making the corresponding disclosure in the return, non-commensurate with such notionally overly stated figure in AIS. Taxpayers are advised to choose the option of ‘information is duplicate’ in such cases.
Similarly, all foreign remittances in LRS are somehow seemingly being perceived as suspicious, and not acknowledging the fact that such remittances may have been made out of fully accounted-for sources.
The problem lies with the fact that AI and ML tools simply compare the value of such remittances or the purchase of immovable property with the returned income and do not recognize that it is not the amount of income but the gross sources of funds, out of which such foreign remittances or purchase of immovable property are being made.
For instance, even if the taxpayer has incurred and reported some capital gain loss from selling securities and mutual funds in its return, the resultant gross sale proceeds may be invested by the taxpayer towards the purchase of immovable property or utilised in making foreign remittances. In such cases, the AI and ML tool will perhaps simply consider the capital loss in the return and ignore the corresponding gross sale proceeds, thereby generating the automated warning message to the taxpayer.
It is time to make artificial intelligence naturally intelligent and machine learning learn by complementing these tools with appropriate human vetting.
Mayank Mohanka is the founder of TaxAaram India and a partner at S M Mohanka & Associates.