30.06.2026
Shubhra Dhanawade
Introduction
The global digital landscape is evolving rapidly. Tax administration has also undergone a significant transformation with the introduction of faceless assessments driven by technology. As part of the digital transition, the Central Board of Direct Taxes vide Notification No. 18/2022 dated 29th March 2022 notified the e-Assessment of Income Escaping Assessment Scheme, 2022 (hereinafter referred to as “the Scheme”) under Section 151A of the Income Tax Act, 1961 (hereinafter referred to as “the Act”).
The scope of the Scheme is to provide assessment, reassessment or recomputation under Section 147 of the Act and issuance of notice under Section 148 of the Act. These functions are carried out through automated allocation. Clause 2(b) of the Scheme defines "automated allocation" as:
“an algorithm for randomised allocation of cases, by using suitable technological tools, including artificial intelligence and machine learning, with a view to optimise the use of resources.”
AI and ML: Are They Independent Technologies
The objective of the provision is clear. Instead of assigning cases manually to officers, cases should be allocated through an algorithmic process. The issue does not arise from the concept of automated allocation itself. Rather, it lies in the legislative language used to describe the technologies enabling that allocation namely, Artificial Intelligence (AI) and Machine Learning (ML).
In common parlance, AI is understood as the broader field of computer science that enables machines to perform tasks requiring human-like intelligence. On the other hand, ML is one of the principal techniques through which artificial intelligence is achieved.
The relationship is commonly illustrated as follows:
Artificial Intelligence
│
├── Machine Learning
│ ├── Deep Learning
│ └── Reinforcement Learning
│
├── Expert Systems
├── Natural Language Processing
└── Computer Vision
From a technical perspective, machine learning does not exist independently of artificial intelligence. Rather, it forms an important subset within the broader field of AI. Therefore, describing AI and ML as distinct technologies may appear conceptually inaccurate.
The OECD AI Principles, the NIST Artificial Intelligence Risk Management Framework, and the European Union AI Act all describe AI broadly while recognising machine learning as one among several computational approaches that enable intelligent systems. These frameworks do not ordinarily treat ML as a separate or independent technological category from AI. This comparative approach demonstrates that legislative terminology can remain technologically accurate while still being sufficiently broad to accommodate future innovation.
A better drafting approach
The concern identified in the foregoing discussion does not undermine the validity of the Scheme. Rather, it highlights an opportunity for more precise legislative drafting.
Instead of referring to:
"artificial intelligence and machine learning"
the provision could have adopted language such as:
"artificial intelligence technologies, including machine learning, deep learning, and other related computational techniques."
Such wording more accurately reflects the relationship between these technologies while preserving the inclusive nature of the definition.
Conclusion
The Scheme represents India’s commitment of technology driven tax administration. However, the reference to AI and ML as separate technological tools raises an important question regarding legislative precision. The future of tax administration will undoubtedly be digital. Ensuring that the law speaks the language of technology with clarity and precision is therefore not simply desirable; it is essential.