Income Tax--Assessment
AI-Generated Tax Orders and the Imperative of Minimal Reasoning : A Constitutional and Legal Perspective
CS. Ayush Rathi
The automation of tax administration in India has brought significant procedural efficiencies. Yet, this transformation also raises a fundamental legal concern: Can a machine-generated tax order devoid of reasons be considered valid? The issue transcends mere operational convenience and engages core constitutional principles, particularly the requirement of natural justice. The Bombay High Court in TPL-HGIEPL Jt. Venture v. Union of India (W.P. (L) No. 15292 of 2024, addressed this question and held that even an AI-generated order must satisfy the minimum requirement of providing reasons, especially where adverse civil consequences follow. This article analyses the ruling within the framework of natural justice and evolving principles of automated tax governance.
1. Introduction : Technology in taxation--Progress with caution
The last decade has witnessed a decisive shift in how tax functions are administered. From the Centralised Processing Centre (CPC) for income-tax returns to faceless assessments and AI-enabled notices, the Indian tax system has embraced digitisation and automation. These reforms are intended to reduce human discretion, improve consistency, and eliminate corruption.
However, the introduction of automated systems into quasi-judicial processes raises an essential legal question: Do orders issued without human involvement satisfy the constitutional mandate of reasoned decision-making? In particular, can a computer-generated tax order that lacks even basic reasoning withstand judicial scrutiny?
2. The principle of reasoned orders : A constitutional expectation
It is a well-established legal doctrine that administrative and quasi-judicial authorities are bound to record reasons when passing adverse orders. This requirement is not merely a statutory obligation but forms part of the broader principle of natural justice, specifically the audi alteram partem rule.