Introduction
Artificial intelligence is no longer a futuristic abstraction for courts: it is a live tool for legal research, case management, transcription and, more controversially, prediction and automation. As artificial intelligence migrates from law-office assistants into courtrooms and state administrative systems, Indian judges and legal actors face a twofold task: first, to re-examine how statutes should be interpreted when artificial intelligence systems shape facts, evidence or the context within which legal norms operate; and second, to calibrate doctrinal principles to protect rule-of-law values such as transparency, accountability and equality against the opaque, probabilistic logic of many artificial intelligence systems.1
Why artificial intelligence matters for statutory interpretation
Statutory interpretation traditionally concerns textualism, purposivism and precedent-based reasoning.2 Artificial intelligence alters the inputs to that process in three ways. First, it changes factual matrices: automated decision systems used by regulators or public bodies can create new categories of administrative action that statutes did not contemplate. Second, artificial intelligence tools influence what judges see and how they research, as machine translation, retrieval and summarisation may surface different authorities or frame an issue in novel ways. Third, predictive analytics can shape expectations about consequences, encouraging courts to consider probabilistic harms rather than discrete legal events. These shifts complicate questions such as whether an existing statute’s scope should be read expansively to cover algorithmic harms, or narrowly to avoid unanticipated regulatory consequences.
Core challenges
A. The black-box problem and the rule of law
Many high-performing artificial intelligence models are opaque: they map inputs to outputs via parameters that are not human-intelligible. When algorithmic outputs are used as evidence, as a basis for administrative action, or as part of judicial decision support, opacity clashes with established procedural norms requiring reasons and intelligibility. Courts must ask whether a statutory provision can be meaningfully applied if the mechanism producing the operative facts is inscrutable, and whether the requirement of reasoned decision-making imposes a de facto right to explanation when artificial intelligence is involved.3
B. Accountability and liability gaps
Statutes allocate rights, duties and liabilities among actors. Automated decision systems often distribute decision-making across software vendors, data providers, public officers and contractors, blurring responsibility. Interpreters of statutes must decide whether liability attaches to the human using the tool, the organisation deploying it, or the tool’s creator, questions that traditional tort and administrative frameworks are ill suited to resolve.
C. Bias, discrimination and evidence integrity
Algorithms trained on historically biased data can reproduce or amplify discrimination. When such systems generate risk scores, classifications or evidence, courts must consider whether existing statutory protections4 for equal treatment, non-discrimination or procedural fairness require fresh interpretive obligations, such as mandatory validation, disclosure of training data, or rigorous evidentiary scrutiny.5
Evolving judicial responses in India
A measured but active approach is emerging in India. The Supreme Court and high courts have not yet produced a canonical doctrine about artificial intelligence and statutory interpretation, but practice and institutional choices signal three trends.
A. Pragmatic adoption coupled with caution
India’s apex court has adopted artificial intelligence for tasks such as translation and transcription, and the court’s digital infrastructure is being enhanced with research tools.6 This pragmatic adoption acknowledges the capacity of artificial intelligence to improve access and efficiency while keeping substantive adjudicative authority firmly human. The posture encourages courts to treat artificial intelligence as an assistive technology, not a substitute for judicial reasoning.
B. Rights-centred scepticism and procedural safeguards
Academic and policy debates in India emphasise a rights-based approach, centring privacy, due process and non-discrimination, when courts confront artificial intelligence-related claims.7 Recent policy pronouncements at the national level and commentary in legal scholarship suggest that Indian courts may demand stronger procedural safeguards, including disclosure, auditing and human oversight, before allowing algorithmic outputs to determine significant rights.8
C. Localised institutional rules
Some high courts and district judicial administrations are moving to govern the use of artificial intelligence internally. A recent example is a policy restricting the use of generative or decision-making artificial intelligence tools by district judiciary officers, emphasising that such tools must not be used for legal reasoning or for replacing human adjudication.9 Such internal policies illustrate judicial conservatism: courts are willing to harness artificial intelligence for administrative tasks but insist on human primacy for legal interpretation and orders.
Interpreting statutes where artificial intelligence creates new harms
How should judges read statutes when artificial intelligence gives rise to novel harms? Three interpretive postures are available.
A. Textual containment
A court could confine statutes to their literal terms, leaving policy adjustments to the legislature. This reduces the risk of judicial policymaking but may permit regulatory gaps where the statutory language was not designed for algorithmic contexts.
B. Purposive extension
Alternatively, courts might interpret statutes purposively, to capture like harms that fall within the legislative aim.10 For example, anti-discrimination or consumer protection statutes might be read to cover algorithmic decisions producing systemic bias or opaque harms.
C. Hybrid doctrines with procedural prerequisites
A third and increasingly attractive option is hybrid: to permit purposive readings that extend statutory reach to algorithmic harms, but to condition such extensions on procedural safeguards such as mandatory disclosure, explainability standards or independent audits, thus balancing remedial reach with rule-of-law norms. This approach preserves democratic legitimacy while providing practical protection.
Practical doctrinal tools for judges
Indian courts can adopt several practical devices to make statutory interpretation resilient to artificial intelligence. Through proceduralisation, they can require that algorithmic decision-making satisfy certain procedures, such as an audit trail and human oversight, before it can ground legal consequences. Through evidentiary filters, they can treat opaque algorithmic outputs as prima facie unreliable unless accompanied by model documentation and validation. Through liability mapping, they can use purposive interpretation to allocate responsibility within statutory regimes, holding deployers accountable where statutes aim to protect affected persons. Finally, through standard-setting remands, where statutory language is ambiguous as to artificial intelligence, they can remand to the regulator with explicit directives to set sectoral standards under delegated legislation,11 preserving the separation of powers while allowing expertise to craft technical norms.
Conclusion: toward a deliberative middle path
Artificial intelligence demands that courts be both imaginative and cautious. Indian judges would do well to preserve core rule-of-law values, namely transparency, reasoned decision-making and equality, while allowing purposive interpretation to close genuine statutory gaps created by algorithmic systems. The most sustainable judicial posture is a deliberative middle path: to permit purposive extensions of protective statutes to algorithmic harms, but to insist on procedural and evidentiary safeguards, including explainability, auditability and human review, that make the governance of artificial intelligence compatible with constitutional commitments. This path respects democratic lawmaking while ensuring that the law remains capable of addressing the novel complexities that artificial intelligence brings to statutory application.
*****
Footnotes
1. For an overview of the incorporation of artificial intelligence into India’s judicial infrastructure, including the e-Courts Phase III framework, see Ministry of Law and Justice, From Digitisation to Intelligence: How AI is Enhancing Access to Justice in India, Press Information Bureau (Govt. of India, May 5, 2025), https://www.pib.gov.in/PressReleasePage.aspx?PRID=2226283.
2. On the mischief rule, the foundational purposive canon directing courts to suppress the mischief and advance the remedy, see Heydon’s Case (1584) 76 Eng. Rep. 637 (Ex.). Indian courts have long favoured purposive construction that reads statutory language in its context rather than in isolation: see Reserve Bank of India v. Peerless General Finance & Investment Co., (1987) 1 SCC 424 (Chinnappa Reddy, J., holding that no word of a statute can be construed in isolation and that a statute must be read to advance its object).
3. On the duty to give reasons as a facet of the rule of law and natural justice, see Kranti Associates (P) Ltd. v. Masood Ahmed Khan, (2010) 9 SCC 496 (holding that quasi-judicial and administrative authorities affecting rights must record intelligible reasons, since reason is the soul of justice and a safeguard against arbitrariness).
4. See India Const. art. 14 (guarantee of equality before the law and equal protection of the laws), the constitutional touchstone against which algorithmic discrimination in state action must be assessed.
5. On the evidentiary threshold for machine-generated output, see Indian Evidence Act, 1872, No. 1 of 1872, s. 65B (special conditions for admissibility of electronic records); now Bharatiya Sakshya Adhiniyam, 2023, No. 47 of 2023, s. 63. On the mandatory certificate requirement, see Anvar P.V. v. P.K. Basheer, (2014) 10 SCC 473, and Arjun Panditrao Khotkar v. Kailash Kushanrao Gorantyal, (2020) 7 SCC 1 (certificate under s. 65B(4) is a condition precedent to admissibility of secondary electronic evidence).
6. On the Supreme Court’s assistive artificial intelligence tools, see Supreme Court of India, Supreme Court Vidhik Anuvaad Software (SUVAS) and Supreme Court Portal for Assistance in Court Efficiency (SUPACE); and generally Ministry of Law and Justice, supra note 1 (e-Courts Phase III research infrastructure).
7. On privacy and informational autonomy as fundamental rights informing scrutiny of algorithmic processing, see K.S. Puttaswamy v. Union of India, (2017) 10 SCC 1 (nine-judge bench recognising the right to privacy under India Const. arts. 14, 19 and 21).
8. On the emerging statutory framework for data-processing safeguards, see the Digital Personal Data Protection Act, 2023, No. 22 of 2023.
9. For the restriction on decision-making use of artificial intelligence by the district judiciary, see High Court of Kerala, Policy Regarding Use of Artificial Intelligence Tools in District Judiciary (July 19, 2025), https://images.assettype.com/theleaflet/2025-07-22/mt4bw6n7/Kerala_HC_AI_Guidelines.pdf.
10. On purposive construction that reads protective statutes to advance their legislative aim rather than their bare letter, see Reserve Bank of India v. Peerless General Finance & Investment Co., supra note 2.
11. On the delegation of technical standard-setting to sectoral regulators under enabling statutes, and its judicial limits, see Hamdard Dawakhana v. Union of India, AIR 1960 SC 554 (subordinate rule-making must be confined by a discernible legislative policy and is void for excessive delegation where the legislature abdicates its essential function).