Introduction
The integration of artificial intelligence into employment decision-making has proceeded at remarkable speed in India’s large corporations. From automated resume screening tools that eliminate applications before a human reviewer sees them, to AI-conducted video interviews that analyse facial expressions and speech patterns to score candidates, to algorithmic performance management systems that recommend termination based on productivity metrics, the technology has penetrated deeply into the employment relationship. Yet the legal framework applicable to these practices remains almost entirely undeveloped.
This is not merely an academic gap. Algorithmic employment tools can encode historical biases in ways that are both powerful and largely invisible. A resume screening algorithm trained on historical hiring data may systematically disadvantage candidates from certain educational institutions, geographic regions, or demographic backgrounds if those patterns are embedded in the training data. A performance management system that uses keylogger data, meeting attendance records, and email response times to score employees may disadvantage workers with disabilities, those with caregiving responsibilities, or those who work in ways that do not fit the productivity model assumed by the algorithm’s designers. The consequences of these algorithmic decisions, unlike the consequences of decisions by individual human managers, are applied at scale and with a consistency that makes them particularly effective instruments of either fair assessment or systematic discrimination.
India’s large IT employers, including Infosys, HCL, Accenture India, and Amazon India’s technology operations, have adopted automated tools across multiple stages of the hiring process. The use of AI in hiring is not limited to the technology sector: banking and financial services, retail, logistics, and manufacturing have all deployed algorithmic hiring tools for high-volume recruitment at entry level. Yet neither the four labour codes, nor the Digital Personal Data Protection Act 2023, nor any sector-specific regulation provides a framework that specifically addresses the fairness, accountability, or transparency obligations of employers using these tools.
Legal Framework
The primary framework governing employment decisions in India continues to be the Industrial Relations Code 2020 and its predecessor, the Industrial Disputes Act 1947, for retrenchment and termination decisions, and the general principles of natural justice for disciplinary proceedings. Neither of these frameworks was designed with algorithmic decision-making in mind, and their application to AI-assisted or AI-recommended employment decisions creates significant interpretive uncertainty.
The principles of natural justice require that before an employee is dismissed, they must be given notice of the charges against them, an opportunity to be heard, and a decision taken by an authority who has applied their mind to the facts. Whether an AI-recommended termination satisfies these requirements depends critically on the degree of human involvement in the final decision. If the AI generates a recommendation and a human manager reviews that recommendation and makes the ultimate decision, the natural justice requirement may arguably be satisfied. If the AI system generates an output that results in automatic termination (as some performance management systems are configured to do), the absence of a reasoning human decision-maker is a serious procedural concern. The question has not been authoritatively decided by an Indian court.
The Digital Personal Data Protection Act 2023 (DPDP Act) is the most recent legislative development with potential relevance to algorithmic employment decisions. The Act grants individuals the right to access personal data processed by data fiduciaries, which would include employment-related data held by employers. An employee who suspects that an AI system has made a discriminatory decision about their employment can, in principle, exercise their right to access the data used in that decision. However, the DPDP Act does not contain any provision equivalent to the General Data Protection Regulation’s Article 22, which grants EU data subjects the right not to be subject to solely automated decisions with significant legal effects, including in the employment context. The DPDP Act’s silence on automated decision-making is a significant omission in the Indian context, given the scale of algorithmic employment decisions being made in the country.
The Equal Remuneration Act 1976, subsumed into the Code on Wages 2019, prohibits discrimination on grounds of gender in matters of remuneration, recruitment, training, transfer, and conditions of service. If an algorithmic hiring tool systematically disadvantages women (or men) relative to candidates of another gender, this statutory prohibition is potentially applicable. However, the remedial mechanism under the Code on Wages is enforcement by a government officer, not private litigation by individual candidates, and the practical accessibility of this remedy for rejected job applicants is limited.
The RPwD Act 2016’s prohibition on discrimination against persons with disabilities applies in the employment context. If an AI hiring tool systematically disadvantages candidates with disabilities, including through the use of tools that assess candidates on criteria that are not relevant to the job but that correlate with disability status (such as speech pattern analysis tools that may disadvantage people with speech impediments), the RPwD Act’s anti-discrimination provisions are engaged.
Judicial Developments
Indian courts have not yet decided a case that directly challenges an AI-based employment decision on discrimination or procedural fairness grounds. However, two streams of doctrine provide building blocks for future litigation.
The first is the natural justice requirement in service law and labour law, which has been consistently enforced by courts as a baseline procedural protection for workers facing dismissal. In a series of Supreme Court decisions, including Workmen v. Firestone Tyre and Rubber Co. of India (1973) and subsequent cases, the Court has held that disciplinary proceedings resulting in dismissal must comply with the principles of natural justice and that an enquiry that is procedurally flawed renders the dismissal invalid. The application of this requirement to AI-assisted disciplinary processes would require courts to determine whether algorithmic recommendations, if made transparent, constitute a sufficient basis for the procedural hearing requirement.
The second stream is the constitutional guarantee of equality under Articles 14 and 16, which applies to public employment. Government employers using AI hiring tools have a constitutional obligation to ensure that those tools do not result in arbitrary or discriminatory outcomes. The application of equality norms to private employment is more contested in Indian constitutional law, but the growing body of doctrine on horizontal application of fundamental rights, and the Supreme Court’s recognition in NALSA v. Union of India (2014) that equality norms inform the interpretation of all legislation including those governing private conduct, provides a basis for extending constitutional accountability to private sector algorithmic employment decisions.
Contemporary Issues and Analysis
Several specific categories of AI employment tool present distinct legal challenges in the Indian context. AI video interview platforms that score candidates on facial expressions and micro-expressions raise concerns about bias against candidates from cultures or communities where the expression norms assumed by the algorithm’s training data do not apply. If an AI interviewing tool was trained primarily on data from Western professional populations, it may systematically undervalue or misinterpret the interview performance of candidates from South Asian or other backgrounds. This would constitute indirect discrimination, even though no discriminatory intent is present.
Automated resume screening tools that filter candidates based on keywords from elite educational institutions or prior employers effectively reproduce existing socioeconomic stratification in the labour market. If a screening algorithm is designed to identify candidates with degrees from IITs, IIMs, and similar institutions, it will systematically disadvantage equally qualified candidates from less prestigious institutions, which in turn disproportionately affects candidates from less affluent socioeconomic backgrounds.
Performance management AI, used by several large Indian employers including in the e-commerce and IT sectors, continuously generates data about employee productivity and uses algorithmic analysis to rank employees, identify underperformers, and generate recommendations for performance improvement plans or termination. The opacity of these systems creates a serious accountability problem: an employee who receives a poor algorithmic performance rating has no mechanism to challenge the criteria or weights applied by the algorithm.
The use of AI in large-scale retrenchment exercises (informally known as “riffs” or “reductions in force” in the tech sector) raises specific concerns under the Industrial Relations Code 2020. If an employer uses algorithmic tools to identify which employees to retrench, the question arises whether the statutory requirement for a retrenchment on the basis of the “last come, first go” principle (seniority-based retrenchment) can be satisfied by an algorithm that applies different criteria.
Comparative and International Perspective
The European Union has taken the most comprehensive regulatory approach to AI in employment. The EU AI Act, which applies from 2024 onwards, classifies AI systems used in recruitment, selection, promotion, termination, task allocation, and performance monitoring as “high-risk AI systems” in Annex III. High-risk AI systems are subject to mandatory conformity assessment before deployment, must meet requirements for transparency, human oversight, data governance, and technical robustness, must be registered in an EU database, and must provide meaningful information to affected individuals. Employers using high-risk AI systems for employment decisions must maintain logs sufficient to enable auditing, and affected individuals must be informed that automated systems are being used.
The United States, at the federal level, has addressed AI in hiring primarily through the framework of existing civil rights law. The Equal Employment Opportunity Commission issued guidance in 2023 clarifying that employers using AI hiring tools are liable under Title VII of the Civil Rights Act 1964 for discriminatory outcomes of those tools, even if the employer did not design the tool and even if the discrimination is unintentional. This disparate impact framework places the burden on the employer to demonstrate that an AI hiring tool that produces discriminatory outcomes is job-related and consistent with business necessity.
New York City has gone further, enacting Local Law 144 in 2023, which requires employers using automated employment decision tools for decisions about candidates or employees in New York City to commission independent bias audits of those tools and to publish the results. The law also requires employers to notify candidates and employees that automated tools are being used and to provide an opportunity to request an alternative selection process.
Practical and Policy Implications
The practical implications of the current legal vacuum are serious. Without transparency requirements, employees and candidates who are adversely affected by algorithmic employment decisions cannot know that an algorithm was involved, cannot access the criteria it applied, and cannot challenge the outcome. Without anti-discrimination testing requirements, employers have no legal incentive to audit their algorithmic tools for discriminatory outcomes. Without human oversight requirements, automated systems can make employment decisions at scale without any individual taking responsibility for those decisions.
Large Indian employers in the technology sector are aware of international regulatory developments, and several have voluntary commitments to fairness and transparency in AI-assisted hiring. However, voluntary commitments in the absence of legal obligations are difficult to verify and may be abandoned under cost or competitive pressure.
The DPDP Act’s data access right, once operationalised through its forthcoming rules, will at least give employees and candidates the ability to request the personal data used in employment-related automated decisions. This is a starting point, but it falls far short of the right not to be subject to solely automated decisions that EU law recognises.
Suggestions and Reforms
India should enact AI-specific employment regulations as subordinate legislation under the DPDP Act or as standalone rules under the Code on Wages, requiring employers who use automated employment decision tools to carry out and publish regular bias audits. The audit requirement should be modelled on New York City’s Local Law 144, adapted to the Indian context.
The DPDP Act should be amended or its rules should specify that where an employer uses automated tools to make or substantially influence employment decisions, the affected individual has the right to an explanation of the criteria applied and the right to request human review of the automated decision. This would bring India’s approach into alignment with the GDPR’s automated decision-making framework.
The EEOC’s disparate impact guidance model should be incorporated into India’s equal remuneration and anti-discrimination framework: employers whose automated hiring or performance management tools produce statistically significant disparities in outcomes across gender, disability status, or other protected characteristics should bear the burden of demonstrating that the tool measures genuine job-relevant capabilities.
The Ministry of Labour and Employment should issue advisory guidelines for employers on ethical AI in hiring and performance management pending the enactment of specific regulations. These guidelines should address transparency, auditability, human oversight, and complaint mechanisms.
Conclusion
Algorithmic employment decision-making is not a future concern; it is a present reality affecting millions of Indian workers and job seekers. The legal framework’s failure to address this reality leaves workers without meaningful recourse and employers without clear guidance on their obligations. The regulatory experience of the EU, the US EEOC, and New York City demonstrates that targeted, proportionate regulation of AI in employment is feasible and effective. India’s position as a major producer and user of AI technology makes the development of a coherent domestic regulatory framework both urgent and achievable. The absence of action is not a neutral policy choice: it is a choice to allow powerful and potentially discriminatory technologies to operate without accountability, at the expense of those least able to challenge their outcomes.
