Harnessing AI for Improved Tax Governance in India

Concept and objectives: Project Insight (PI), launched by the Income Tax Department (ITD), is an AI-driven initiative aimed at improving tax compliance, transparency, and revenue mobilisation. It addresses India's structural challenge of a low tax-GDP ratio (around 16.36%) and significant tax evasion losses. The project seeks to shift tax administration from a reactive enforcement model to a data-driven, predictive, and preventive system.

Key components: PI operates through three major pillars:

• INTRAC (Income Tax Transaction Analysis Centre): Uses AI and big data to create a 360-degree financial profile of taxpayers by integrating data from banking, GST, securities, and high-value transactions.
• Compliance Management System: Identifies discrepancies and prioritizes cases for action.
• NUDGE strategy: Encourages voluntary compliance by sending reminders to taxpayers to correct mismatches.

Addressing structural weaknesses: India’s low tax-GDP ratio and high levels of tax evasion undermine its fiscal capacity. AI-driven systems like Project Insight help identify hidden income patterns and discrepancies, thereby increasing revenue without raising tax rates. This is crucial for funding public goods such as infrastructure, health, and education.

Improving efficiency and compliance: AI enables risk-based targeting, allowing tax authorities to focus on high-value evasion cases rather than random audits. The NUDGE approach promotes voluntary compliance, reducing litigation and administrative burden. For instance, over one crore revised returns have been filed, generating additional revenue of ₹11,000 crore.

Global competitiveness: Advanced economies like the U.S., U.K., and Australia have already adopted AI in tax systems. For India to remain competitive and improve ease of doing business, modernizing tax administration is essential. Thus, AI adoption is not just a technological upgrade but a strategic fiscal reform.

Data integration and analytics: Traditional tax systems relied heavily on manual audits and self-reported data. AI systems like PI integrate data from multiple sources to create comprehensive taxpayer profiles. This allows detection of inconsistencies between declared income and actual financial behaviour.

Behavioural interventions: Instead of punitive action, AI enables proactive compliance through nudges. Taxpayers are alerted to discrepancies and given an opportunity to correct them voluntarily. This reduces adversarial interactions and builds trust. For example, targeted campaigns have led to significant corrections in foreign asset declarations and deduction claims.

Operational efficiency: AI automates routine tasks such as data processing and risk assessment, freeing up human resources for complex decision-making. It also reduces processing time, as seen in the reduction of refund timelines from 93 days to 17 days. Thus, AI enhances both effectiveness and efficiency in tax administration.

Data quality and false positives: AI systems depend heavily on the quality of input data. Inaccurate or incomplete data can lead to false identification of tax evasion. For instance, legitimate financial behaviour, such as joint family transactions or use of prior savings, may be flagged as suspicious.

Algorithmic bias and fairness: AI models trained on historical data may replicate existing biases, disproportionately targeting certain regions or socio-economic groups. The Dutch childcare benefits scandal is a prominent example where biased algorithms led to wrongful accusations.

Privacy and accountability concerns: AI systems require access to sensitive financial data, raising concerns about data security and misuse. Additionally, lack of transparency in decision-making processes can undermine trust. Without mechanisms for human oversight and grievance redressal, such systems risk becoming opaque and unaccountable.

Efficiency gains: AI-driven systems significantly improve efficiency by automating data analysis, identifying high-risk cases, and reducing administrative delays. They also promote voluntary compliance through behavioural nudges, making tax enforcement less intrusive. This enhances revenue collection and reduces the burden on tax authorities.

Challenges to fairness: However, efficiency gains may come at the cost of fairness and due process. Algorithmic bias, lack of explainability, and false positives can lead to unjust outcomes. Taxpayers may find it difficult to challenge decisions if the system lacks transparency. This raises concerns about procedural justice.

Need for safeguards: To balance efficiency with fairness, robust governance mechanisms are required, including human-in-the-loop decision-making, independent audits, and clear grievance redressal systems. Without such safeguards, AI-driven systems risk undermining trust in the tax system. Thus, while AI offers transformative potential, its deployment must be ethically grounded and legally accountable.

Improved voluntary compliance: One of the most notable outcomes of Project Insight is the increase in voluntary compliance. Over one crore taxpayers have filed revised returns since 2020-21, leading to additional tax revenues of ₹11,000 crore. This demonstrates the effectiveness of the NUDGE strategy in influencing taxpayer behaviour.

Detection of large-scale evasion: AI tools have enabled the identification of sophisticated tax evasion methods. For instance, the ITD uncovered ₹70,000 crore worth of suppressed sales in the restaurant sector through data analytics. This highlights the system’s ability to detect complex fraud patterns that traditional methods might miss.

Enhanced service delivery: The reduction in tax refund processing time from 93 days to 17 days reflects improved administrative efficiency. Additionally, targeted campaigns have led to increased disclosure of foreign assets and correction of false claims. These outcomes illustrate how AI can simultaneously improve compliance, enforcement, and taxpayer experience.

Institutional safeguards: Establishing an independent AI ombudsperson is crucial to handle grievances and review contested decisions. Regular algorithmic audits and impact assessments should be mandated to ensure fairness and transparency. Public reporting of false positives and appeal success rates can enhance accountability.

Legal and technical measures: Strong data protection laws must be enforced to safeguard taxpayer information. AI systems should be designed with explainability features, allowing taxpayers to understand how decisions are made. Incorporating human oversight in high-stakes decisions ensures that technology does not override due process.

Policy approach: A balanced approach combining innovation with ethics is essential. Policymakers should promote capacity building, stakeholder consultation, and phased implementation. By aligning technological advancement with constitutional principles, India can build a tax system that is both efficient and trustworthy.