Introduction: Law in Society and a new socio-legal challenge
According to the sociological school of law, law is not merely something that comes from society; it should also work to reduce unfairness and inequality in society. Roscoe Pound called this role social engineering, by which he meant the management of conflicting interests. The Law in Society approach examines the difference between how laws are written and how they actually operate, especially when technology gives rise to new kinds of social problems.1 AI-based recruitment is a good example of this kind of modern socio-legal issue.
Many companies in India now use AI tools such as HireVue, Pymetrics, LinkedIn Talent Insights and other Applicant Tracking System (ATS) software to help with hiring. While the makers and users of these tools say that they remove human bias,2 research shows that tools trained on past hiring data learn the patterns in that data and so reproduce the prejudices it records.3 In one widely reported case, an experimental tool penalised resumes that included the word ‘women’s’, as in ‘women’s chess club captain’.4 The same mechanism could equally lower the scores of resumes that show career breaks related to maternity, or that carry surnames associated with particular castes.
Thus, a technology meant to ensure fairness has instead become a cause of unfairness. This runs against the promise of Article 14, which guarantees everyone equality before the law and the equal protection of the laws.
The problem is not only technological; it is also social and legal. There is a conflict between an employer’s right to carry on a business under Article 19(1)(g) and a job seeker’s right to equal opportunity, as protected by Articles 14, 15 and 16 of the Constitution.5 This conflict lies at the heart of the theme ‘Law in Society: Contemporary Socio-Legal Challenges and Emerging Legal Responses’.
Research questions
1. Why should AI-based hiring bias be treated, from a Law in Society point of view, as a socio-legal issue rather than merely a technical problem?
2. Are the current Indian laws adequate to control algorithmic discrimination?
3. What practical difficulties in ensuring fair AI hiring create a gap between the law on paper and the law as applied?
4. What new legal responses are emerging from courts and from other jurisdictions to deal with this socio-legal problem?
Objectives
1. To examine AI recruitment bias as a modern socio-legal issue using the Law in Society approach.
2. To examine closely the laws that deal with equality and identify where they are missing or inadequate.
3. To identify the practical problems that make the law on paper differ from the law in practice.
4. To study new legal ideas and international instruments that are being used in response to these issues.
5. To suggest legal changes that can make the law in India fairer and more responsive to society.
Research methodology: purely doctrinal
The research is based entirely on the study of laws and legal principles. There is no survey or testing of AI software. The main sources are Articles 14, 15, 16, 19(1)(g) and 21 of the Constitution of India, the Information Technology Act, 2000,6 the Digital Personal Data Protection Act, 2023,7 the Equal Remuneration Act, 1976 (now repealed and replaced by the Code on Wages, 2019),8 and the Rights of Persons with Disabilities Act, 2016.9 Supreme Court decisions such as K.S. Puttaswamy v. Union of India,10 State of West Bengal v. Anwar Ali Sarkar,11 Maneka Gandhi v. Union of India12 and Indra Sawhney v. Union of India13 have also been used. Other sources, such as the NITI Aayog National Strategy for Artificial Intelligence (2018),14 the EU AI Act (2024)15 and the UNESCO Recommendation on the Ethics of Artificial Intelligence (2021),16 are used to understand how other parts of the world are responding to these issues.
AI recruitment bias as a contemporary socio-legal challenge
A. Breaking the idea of real equality under Article 14
Article 14 forbids arbitrary State action; as the Supreme Court put it in E.P. Royappa v. State of Tamil Nadu, equality is antithetic to arbitrariness.17 AI hiring tools that score candidates in this way would not pass the test of reasonable classification applied in State of West Bengal v. Anwar Ali Sarkar, which requires an intelligible differentia having a rational relation to the object sought to be achieved.18 If AI gives lower scores on the basis of gender, caste or disability, there is no clear reason for that decision and no logical connection to job performance. A well-known example is the experimental resume-screening tool that Amazon began building in 2014, which penalised resumes containing the word ‘women’s’; the company disbanded the team behind it by the start of 2017.19 This shows how old social biases are built into technology, creating new social and legal challenges.
B. Intersectional discrimination under Articles 15 and 16
Discrimination by AI is not of one type; it operates on several grounds at the same time. A Dalit woman who takes a break from her career because of pregnancy, for example, faces compounded disadvantage. Article 15 prohibits discrimination on grounds including sex and caste, while Article 16 guarantees equality of opportunity in public employment.20 Even though private employment falls outside Article 16, the spirit of Article 15 and the Directive Principles of State Policy make the State responsible for ensuring equal treatment for everyone. The Law in Society approach shows that technology which appears neutral can in fact continue unfair treatment in society.
C. Violation of privacy and dignity under Article 21
Many AI tools analyse facial features, tone of voice and even social media activity. In K.S. Puttaswamy v. Union of India, the Supreme Court recognised informational privacy as a facet of the right to privacy under Article 21, and Kaul J., concurring, warned that profiling can result in discrimination based on religion, ethnicity and caste.21 In AI hiring, job seekers are often not told that their facial expressions or voice are being analysed, which affects their dignity. This is a serious challenge under Article 21.
D. Structural risk for future generations
If AI keeps learning from old data that is unfair, it will carry that unfairness forward to future generations. That runs against the idea of intergenerational fairness and long-term justice. The NITI Aayog’s National Strategy for Artificial Intelligence (2018) also warned that the existing data on which AI systems are built ‘may have biases, which may have got reinforced over time’.22
Analysis of the present legal framework: law in books
A. Constitutional base
The Indian Constitution provides a strong foundation. Article 14 guarantees equality before the law. Article 15 prohibits discrimination. Article 16 ensures equality of opportunity in public employment. Article 19(1)(g) protects the freedom to carry on any occupation, trade or business, subject to reasonable restrictions. Article 21 guarantees the right to life with dignity, which, it is argued here, includes protection against unfair AI decisions.
B. Statutory framework: fragmented and silent
1. Information Technology Act, 2000: Section 43A and the Information Technology (Reasonable Security Practices and Procedures and Sensitive Personal Data or Information) Rules, 2011 make a body corporate that is negligent in protecting sensitive personal data liable to pay compensation.23 However, there is no rule on AI bias, no right to know why a decision was made, and no mandatory audit. Section 66A of the Act was struck down in Shreya Singhal v. Union of India.24
2. Digital Personal Data Protection Act, 2023: This Act is important but not sufficient for AI bias. It defines the Data Fiduciary, requires consent, and sets rules for the processing of personal data.25 However, it does not include (a) a right to an explanation of decisions made by AI, unlike the EU’s General Data Protection Regulation,26 (b) a requirement to assess the impact of AI, or (c) a prohibition on AI inferring attributes such as caste; and (d) it allows personal data to be processed without consent ‘for the purposes of employment’.27
3. Equal Remuneration Act, 1976 and Code on Wages, 2019: The Code on Wages, whose remaining provisions were brought into force on 21 November 2025, repealed and replaced the Equal Remuneration Act. Both prohibit discrimination only on the ground of sex (in the Code, gender): in wages for the same or similar work and, for such work, in recruitment.28 They do not cover caste, disability or AI-related indirect discrimination.
4. Rights of Persons with Disabilities Act, 2016: Section 20 prohibits Government establishments from discriminating against persons with disabilities in any matter relating to employment,29 but there is no rule for AI tools that could reject people with speech disabilities because of voice analysis.
C. Institutional framework
The Ministry of Electronics and Information Technology frames AI policy and NITI Aayog has a National Strategy for Artificial Intelligence, but these are only guidelines and are not enforceable. There is no dedicated body such as the European Union’s AI Office,30 and no mandatory audit. This shows a gap between the rules that exist and what society needs.
Practical challenges: law in books versus law in action, the core of Law in Society
1. Black box and invisibility problem: To prove discrimination under Article 14 there must be evidence, but AI algorithms are kept secret as trade secrets. Candidates do not know the reasons for their rejection, and without a right to know the reason it is hard to take action. This shows the gap between what is written in law and how it works in practice.
2. Lack of data and proof: To show bias under Article 14, statistical data about hiring outcomes is needed. But this data stays with the employer and is not shared publicly.
3. No standard for fairness: India has no official rule like the US four-fifths (80%) rule, under which a selection rate for any group that is less than four-fifths of the rate for the group with the highest rate is generally regarded as evidence of adverse impact.31 Without such a rule in the DPDP Act or elsewhere, it is hard to enforce fairness.
4. Financial problems for small businesses: Large companies can afford to have their tools audited for bias. Most employers in India, however, are MSMEs that use inexpensive ATS software without any check for bias. A common audit facility is needed, on the model of the common effluent treatment plants used in environmental law.
5. Private sector and Article 14: Article 14 applies directly to the State, but where there is no law and fundamental rights are being violated, the Supreme Court can lay down guidelines that bind private employers as well, as it did in Vishaka v. State of Rajasthan.32
6. No regulator or grievance system: The DPDP Act establishes a Data Protection Board, but it does not require any of its members to have expertise in algorithmic fairness.33 There is also no fast remedy for job seekers who lose out on opportunities.
7. Lack of awareness: In places such as Sangli and Kolhapur, job seekers may not know that their resumes are screened by AI, and HR managers may treat AI scores as final without checking them again. This shows that the law is not yet part of everyday life.
Emerging legal responses: law as an instrument of social change
Where there is no specific law, constitutional principles and global models can help. This is the central idea of Law in Society.
1. Doctrine of non-arbitrariness under Article 14: In Maneka Gandhi v. Union of India, the Supreme Court held that Article 14 strikes at arbitrariness and that a procedure must be right, just and fair, not arbitrary, fanciful or oppressive.34 AI decisions that give no reasons are arbitrary and contrary to this standard.
2. Right to privacy and informational self-determination: In K.S. Puttaswamy, the Supreme Court recognised informational privacy, and Kaul J., concurring, warned against profiling and discussed a qualified right to be forgotten.35 This can support the argument that people should not be judged by AI alone in hiring.
3. Reasonable classification test: AI-based classification should rest only on job-related factors, not on factors such as names, addresses or career breaks.
4. Global emerging response, the EU AI Act, 2024: The European Union classifies AI used in recruitment and selection as high-risk and requires such systems to be subject to risk management, to examination for possible biases, to transparency and to human oversight.36 This is a model India can follow.
5. US and UK responses: In 2023 the US Equal Employment Opportunity Commission issued technical assistance stating that AI selection tools are subject to Title VII of the Civil Rights Act of 1964 and that an employer may be liable for their disparate impact even where no discrimination was intended; the document was removed from the Commission’s website in January 2025.37 The UK’s Equality Act 2010 prohibits indirect discrimination, and its protection of job applicants extends to the arrangements an employer makes for deciding to whom to offer employment, including arrangements that use AI.38
Suggestions for a socially responsive legal framework
1. Amend the DPDP Act, 2023: Add a special list of high-risk AI and include recruitment AI in that list, as the EU AI Act does. Also require an algorithmic impact assessment before such a system is used.
2. Right to know and human review: Add a new Section 11A to the DPDP Act giving people the right to know why an automated decision was made and the right to have a human review it, following the Puttaswamy decision and the principle of fairness.
3. Mandatory bias audit through a common facility: Set up a common audit centre for small and medium businesses through NASSCOM or STPI, to check algorithms for bias in a way that is affordable and fair.
4. Codify the disparate impact rule: Add penalties that depend on the size of the company and how often it breaks the rules, and write a rule like the US four-fifths (80%) rule into law.
5. Transparency: Require job portals such as Naukri and LinkedIn to disclose that AI is being used in the hiring process, and require companies to publish diversity figures on their websites every quarter.
6. Mandatory training: Make training approved by MeitY mandatory for HR managers, and include AI ethics in MBA and HR courses.
7. Create an AI equality office: Set up an independent body, similar to the Data Protection Board, with technical experts to oversee AI used in hiring and to keep the law in step with society’s needs.
Conclusion
Seen through the lens of society, bias in AI hiring is more than a technical mistake. It is a new legal and social problem arising from the conflict between business efficiency under Article 19(1)(g) and equal treatment under Articles 14, 15, 16 and 21. This study shows that while the courts have developed protections for equality, privacy and fairness through decisions such as Puttaswamy, Maneka Gandhi, Royappa and Vishaka, the data protection laws, such as the IT Act, 2000 and the DPDP Act, 2023, have still not addressed algorithmic discrimination. This gap between law on paper and law in practice is a key concern of the Law in Society movement. A sound legal system must treat algorithmic bias as legal discrimination, recognise a right to know, require audits and provide for human review. Only then can the law properly support equality and social justice, which is the main goal of the Law in Society theme.
*****
Footnotes
1. See generally Lawrence M. Friedman, Law and Society: An Introduction (1977); Roger Cotterrell, Law, Culture and Society: Legal Ideas in the Mirror of Social Theory (2006).
2. See Ifeoma Ajunwa, The Paradox of Automation as Anti-Bias Intervention, 41 Cardozo L. Rev. 1671 (2020) (arguing that automated hiring, adopted as an anti-bias intervention, has in some instances replicated and amplified bias).
3. Solon Barocas & Andrew D. Selbst, Big Data’s Disparate Impact, 104 Calif. L. Rev. 671 (2016).
4. Jeffrey Dastin, Amazon Scraps Secret AI Recruiting Tool That Showed Bias Against Women, Reuters (Oct. 10, 2018), https://www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G.
5. India Const. arts. 14, 15, 16, 19(1)(g), 21.
6. The Information Technology Act, No. 21 of 2000, India Code (2000).
7. The Digital Personal Data Protection Act, No. 22 of 2023, India Code (2023). Sections 18 to 26, on the Data Protection Board, came into force on Nov. 13, 2025, when the Digital Personal Data Protection Rules, 2025 were notified; most of the remaining provisions, including §§ 3 to 17 (except § 6(9), which comes into force one year after that notification), come into force eighteen months after it.
8. The Equal Remuneration Act, No. 25 of 1976, India Code (1976), repealed by The Code on Wages, No. 29 of 2019, India Code (2019), § 69 (repeal effective Nov. 21, 2025).
9. The Rights of Persons with Disabilities Act, No. 49 of 2016, India Code (2016).
10. Justice K.S. Puttaswamy (Retd.) v. Union of India, (2017) 10 SCC 1 (India).
11. State of West Bengal v. Anwar Ali Sarkar, AIR 1952 SC 75 (India).
12. Maneka Gandhi v. Union of India, AIR 1978 SC 597 (India).
13. Indra Sawhney v. Union of India, AIR 1993 SC 477 (India).
14. NITI Aayog, National Strategy for Artificial Intelligence #AIForAll (Discussion Paper, 2018), https://www.niti.gov.in/sites/default/files/2023-03/National-Strategy-for-Artificial-Intelligence.pdf.
15. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act), 2024 O.J. (L 2024/1689) [hereinafter EU AI Act].
16. UNESCO, Recommendation on the Ethics of Artificial Intelligence (Nov. 23, 2021).
17. E.P. Royappa v. State of Tamil Nadu, (1974) 4 SCC 3 (India).
18. Anwar Ali Sarkar, AIR 1952 SC 75.
19. Dastin, supra note 4.
20. India Const. arts. 15(1), 16(1)–(2).
21. Puttaswamy, (2017) 10 SCC 1 (recognising informational privacy as a facet of the right to privacy); id. (Kaul, J., concurring) (observing that profiling “can result in discrimination based on religion, ethnicity and caste”).
22. NITI Aayog, supra note 14, at 85.
23. Information Technology Act § 43A; Information Technology (Reasonable Security Practices and Procedures and Sensitive Personal Data or Information) Rules, 2011, G.S.R. 313(E) (Apr. 11, 2011) (India). Section 43A is omitted by § 44(2) of the Digital Personal Data Protection Act, 2023, a provision in the group that comes into force eighteen months after Nov. 13, 2025.
24. Shreya Singhal v. Union of India, (2015) 5 SCC 1 (India).
25. Digital Personal Data Protection Act §§ 2(i), 4, 6.
26. Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data and on the Free Movement of Such Data (General Data Protection Regulation), arts. 13(2)(f), 15(1)(h), 22, 2016 O.J. (L 119) 1.
27. Digital Personal Data Protection Act § 7(i).
28. Equal Remuneration Act §§ 4, 5; Code on Wages §§ 3, 69.
29. Rights of Persons with Disabilities Act § 20(1).
30. EU AI Act, supra note 15, art. 64.
31. Uniform Guidelines on Employee Selection Procedures, 29 C.F.R. § 1607.4(D) (2025).
32. Vishaka v. State of Rajasthan, AIR 1997 SC 3011 (India).
33. Digital Personal Data Protection Act §§ 18, 19(3).
34. Maneka Gandhi, AIR 1978 SC 597.
35. Puttaswamy, (2017) 10 SCC 1 (Kaul, J., concurring).
36. EU AI Act, supra note 15, arts. 6(2), 9, 10(2)(f)–(g), 13, 14 & annex III, point 4(a).
37. U.S. Equal Emp. Opportunity Comm’n, EEOC-NVTA-2023-2, Select Issues: Assessing Adverse Impact in Software, Algorithms, and Artificial Intelligence Used in Employment Selection Procedures Under Title VII of the Civil Rights Act of 1964 (May 18, 2023) (technical assistance document, removed from the Commission’s website on Jan. 27, 2025).
38. Equality Act 2010, c. 15, §§ 19, 39(1)(a) (UK).