AI-Powered Tax Governance in India

Context

Artificial Intelligence (AI) is changing the system of tax administration in India, where it used to be based on a traditional, manual, and often confrontational system to a non-intrusive, data-driven and efficient system. India has a low tax-to-GDP ratio (ca. 16.36% in 2021-22) and a high rate of tax evasion, which means that the introduction of AI is of paramount importance to enhance the fiscal capacity.

AI Applications in Indian Tax Administration

The leaders in this digital change are the Income Tax Department (ITD) and the Central Board of Indirect Taxes and Customs (CBIC):

• Project Insight (PI): It will be established in 2017 and leverage AI and machine learning to create a 360-degree profile of taxpayers.
• INTRAC (Income Tax Transaction Analysis Centre): It is the AI analysis machine that process data about the banking, securities, property dealings, and GSTN to identify inconsistencies.
• NUDGE Strategy (Behavioral Insights): AI-based nudges can send both SMS/email messages to taxpayers to enable them to amend returns of their choice, which has also enhanced compliance without applied force.
• Faceless Assessment and Appeals: The AI will select and allocate cases without human intervention, which will minimise corruption and human judgment.
• GSTN Analytics: AI processes are used to match invoices between GSTR-1, GSTR-2B and GSTR-3B to detect fake ITC claims and circular trading networks.
• Chatbots/Virtual Assistants: Such systems as ‘Kar Saathi’ (open in 2026) will be able to help taxpayers 24/7.

Benefits and Measurable Outcomes

• Higher Revenue and Higher compliance: The Nudge campaigns by AI have resulted in more than 11,000 crore of revenue and millions of taxpayers have filed corrected returns.
• Identification of Large-Scale Fraud: AI has been used to identify huge evasion such as 70,000 crore of hidden turnover in the restaurant industry.
• Improved Taxpayer Service: It has improved the time by 93 days to about 17 days on the time it takes to process the refund.
• Effective Implementation: Allows targeted audits and not arbitrary inspections, which are risk-driven.

Challenges and Risks in AI-Based Tax Governance

Despite its potential, AI adoption faces significant hurdles:

1. Problems with Data Quality & False Positives: AI systems can misunderstand complicated, yet legal financial activity and treat the purchase of real transactions as evasion and result in the needless generation of compliance costs.
2. Algorithmic Bias: AI models that are trained on biased historical data can unwillingly criticize some social-economic or geographic populations, which can result in fairness problems.
3. Absence of Transparency (Black Box Problem): The taxpayer might not know the reasons as to why they are flagged, which should not be trusted, and it makes it difficult to argue with the decisions.
4. Privacy and Security Risk: Pooling so much individual financial information into one system offers a huge cyber attack and abuse target.
5. Institutional Gaps: A lack of a special AI ombudsman or independent control of algorithms prevents responsibility.

The Way Forward

To make AI-driven tax governance effective and equitable, the following measures are needed:

• Human-in-the-Loop: AI must not take the place of a judge, but it must support. Important decisions can not be made without human checking.
• Explainable AI (XAI): Tax authorities should make sure that AI models can be transparent in the process of arriving at decisions.
• Excellent Data Governance: Linking AI tools to the Data Protection Act in order to guarantee the privacy of taxpayers.
• Creating Accountability Systems: The creation of an AI ombudsperson to evaluate disputed rulings.

Conclusion

AI is transforming the tax environment in India, providing a way to a new compliance level and better revenue collection. Nevertheless, to protect the interests of an effective Viksit Bharat, such a technology should have strong ethical protective measures that are able to maintain fairness, transparency and trust and turn this mechanism into a citizen-centric, non-invasive, and just one.

Source: The Hindu

Download as PDF