Introduction
As state agencies adopt artificial intelligence algorithms to make decisions in public employment, the distribution of welfare benefits, predictive policing, and criminal sentencing, this raises an essential constitutional question. Algorithmic tools are typically presented as neutral, objective, and cost-effective replacements for human judgment. Yet, machine learning algorithms often perpetuate and magnify existing biases built into society.
Given that an algorithm trained on poorly collected or inherently discriminatory historical data might lead to biased results that unfairly harm protected social groups, legal experts and the Constitutional courts are faced with an important question: Could an algorithmic decision lead to a violation of Article 14 of the Constitution of India?
Understanding Article 14: The Dual Doctrine of Equality
Article 14 provides for two fundamental tests and mentions them below,Equality before law.No citizen of India shall be denied equality before the law or the equal protection of the laws within the territory of India Article 14 guarantees to all persons within the territory of India, “equality before law” and the “equal protection of the laws”.
1. The Doctrine of Reasonable Classification
In order to meet constitutional muster, a classification or distinction established by the State must pass muster under two-pronged test:
- Intelligible Differentia: The class must have been based on some real and discernible difference, or some difference among members of the group.
- Reasonable Connection: There must exist a reasonable and direct connection to the very object aimed at being brought about by the legislation or executive order.
2. The Doctrine of Non-Arbitrariness
In E.P. Royappa v. State of Tamil Nadu (1974) and reiterated in Maneka Gandhi (1978), the court opined that”equality is antithetical to arbitrariness”. Any State action which is unreasonable, discriminatory, irrational or clearly arbitrary in nature would infringe Article 14.
Anatomy of Algorithmic Bias: How AI Discriminates
Algorithmic systems are not autonomous, they mirror the dataset with which they were trained, and the parameters set by the engineers who created them:
1.Historical Trends Bias:
A predictive police algorithm is biased against some communities or neighborhoods if the historical data used to feed the algorithm contains biased policing patterns that were based on discriminatory policing strategies; so, predictions will repeat these patterns.
2. Proxy Discrimination:
While protected attributes (like caste, religion and gender) may be not be explicitly included in the data of an algorithm, machine learning algorithms can readily find proxy variables (like postal codes, income brackets, and educational background).
3. “Black Box” Issue:
Deep neural networks and other sophisticated ML algorithms are often black boxes. If an AI model turns down a welfare application or marks a job candidate “high risk”, it can be challenging for humans to figure out exactly why it was denied.
How AI Violations Trigger Article 14
Automated decision-making algorithms and transparency triggers arise whenever state institutions enact such algorithms that are hindered by algorithmic bias and opaque features and an award of these algorithms has a substantial impact on a person’s rights so as to cause a constitutional violation under Article 14 in the areas of :
1. Failure of Intelligible Differentia and Rational Nexus
If the system is programmed to classify people or exclude the public based on attributes (proxy variables) that have no actual nexus to merit or entitlement, the system is devoid of an intelligible differentia.
Similarly, deploying an AI scoring algorithm in the public employment process that disadvantages candidates based on spatially Referenced Historical Data set, would be an arbitrary criterion, and does not have a rational nexus to the task, and would therefore be unconstitutional.
2. Manifest Arbitrariness and Opaque Decision-Making
Under the non-arbitrariness doctrine, executive decisions that impact individual rights should be reasoned and explicable. A black box decision-relying on a opaque algorithm that withholds social welfare benefits from an individual citizen-would constitute an obviously arbitrary action by the state.
Without the ability to do anything about an unreasoned computer-generated denial, the state’s use of these algorithms takes away essential constitutional rights.
3. Indirect Discrimination and Disparate Impact
In Indian constitutional law, however, there has been a shift to recognize indirect discrimination, as in seminal privacy cases such as the Justice K.S. Puttaswamy v. Union of India (2017) case, and increasingly in courts’ anti-discrimination jurisprudence.
Similarly, even if an algorithm does not have a discriminatory intent, if its implementation results in severe harm to a protected class, then it violates equal protection.
Strategic Remedies for Constitutional AI Compliance
In addition to structural safeguards within the legal and technological infrastructure to prevent Article 14 violations by state-level automated systems:
1. AIAs:
State agencies shall require automated tools in public decision-making to be subjected to biennial impact assessments by the state that analyze the potential for bias, proxy, and otherwise risks, including all compliance with state and federal constitutional rights.
2. The “Right to Explanation” and Algorithmic Auditing:
We have to ban black-box systems on which states are relying. Automated decisions impacting fundamental rights should be accompanied by human-understandable explanations, so people can understand how the decision was taken.
3. Implement Human-in-the-Loop Safeguards:
No matter how automated the system, it must be used as a guidance or decision support tool, not as a substitute for human judgment. Critical decisions on civil liberties, public employment, or welfare rights should be subject to human review.
Conclusion
Can AI breach Article 14? Yes, without any doubt. Any state decision which transfers admin discretion to automated algorithm is no less an exercise of State power under the Constitution. An algorithm which indirectly causes discrimination via proxy variables, or which denies citizens public benefits through unaccountable processes, is an arbitrary and unlawful use of State power.
However, automation by the state should not be used as an excuse to undermine the guarantee of equality before the law. As governments increasingly automate functions, honouring the promise of Article 14 will require ensuring that algorithms are subject to the same constitutional safeguards of fairness, transparency and non-discrimination as human states.
References
2. Maneka Gandhi v. Union of India, (1978) 1 SCC 248: Affirmed that state procedure must be just, fair, and non-arbitrary to satisfy constitutional guarantees of liberty and equality.
3. Justice K.S. Puttaswamy (Retd.) v. Union of India, (2017) 10 SCC 1: Established that fundamental rights must be interpreted dynamically to protect personal autonomy and dignity against modern technological erosion.
4. Shreyas, A. & Narayanan, R. (2025/2026): Automated Governance and the Indian Constitution: Testing AI Decision-Making Against Article 14, Indian Journal of Constitutional Law, examining judicial standards for algorithmic accountability.
5. Constitution of India: Article 14 (Equality Before Law) and Article 12 (Definition of State Agencies bound by Part III Fundamental Rights).
