The justice delivery system in India, long plagued by case backlog and procedural rigidity, is undergoing a digital overhaul. Central to this transformation is the Artificial Intelligence in Judicial Decision-Making Act, 2023 (AIJDA 2023), which marks a watershed moment in the state’s approach to justice through algorithmic assistance. By institutionalizing the AI-driven JusPredict system, the Act promises efficiency, consistency, and objectivity in judicial outcomes, especially in criminal adjudication. However, this integration has sparked deep constitutional debates over judicial discretion, fundamental rights, and data privacy.

Conceptual Framework: The JusPredict System

JusPredict is an AI-powered risk assessment tool trained on a wide corpus of judicial and extrajudicial data—from judgments delivered between 2000 to 2022 to behavioral and social metadata, including social media patterns and geolocation information. Its central output is a “risk score”, categorizing an accused’s potential threat to public safety, reoffending likelihood, or absconding risk. This score significantly influences pre-trial bail and detention decisions.

While designed for “data-driven justice,” JusPredict’s algorithm is not open to public scrutiny due to its protection under trade secrecy laws, especially under the Information Technology Act, 2000. Consequently, judges, particularly at the lower judiciary level, often feel compelled to defer to the AI’s outputs to align with emerging institutional norms, risking judicial independence.

Case Background and Emerging Concerns

Two recent high-profile cases underscore the real-world consequences of algorithmic justice:

• Aarav Kumar, a 29-year-old social activist, was denied bail despite no direct evidence of criminal conduct. His high-risk score derived from mere presence at protest sites and digital communications was cited as the key basis for denial of bail.
• Meera Singh, a blogger, was similarly detained based on AI interpretation of blog metadata, associations, and hashtags, despite no legal proof of extremist intent or prior convictions.

In both instances, the deprivation of liberty was grounded not in legal certainty or evidentiary proof, but in probabilistic AI forecasting. These cases raise urgent concerns about due process and potential violations of Article 14 (Right to Equality) and Article 21 (Right to Life and Personal Liberty) of the Indian Constitution.

I. Constitutional Challenges Posed by Algorithmic Judicial Systems

1. Violation of Article 14: Arbitrary and Unequal Treatment

Article 14 guarantees equality before law and prohibits arbitrary state action. However, the JusPredict system, by functioning as a “black box,” potentially entrenches and automates historical biases. If historical data includes discriminatory trends based on caste, religion, or socio-economic status, the AI will learn and perpetuate these patterns. This algorithmic opacity and inability of an accused to challenge the rationale behind a risk score could constitute an arbitrary classification lacking any reasonable nexus, thus violating Article 14.

2. Violation of Article 21: Due Process and Liberty

The use of JusPredict without providing the accused an opportunity to rebut the algorithmic assessment undermines the right to a fair hearing—an essential component of Article 21, as recognized in Maneka Gandhi v. Union of India. Furthermore, the system’s opaque logic may deprive individuals of liberty without procedural safeguards, a core aspect of substantive and procedural due process jurisprudence.

II. Data Privacy and Informational Autonomy

The AIJDA 2023’s use of personal, behavioral, and location data raises serious privacy issues under the Digital Personal Data Protection Act, 2023 and the Supreme Court’s ruling in K.S. Puttaswamy v. Union of India, which recognized the right to privacy as a fundamental right under Article 21.

Key Concerns:

• Lack of Consent: Individuals are neither informed nor do they provide consent for the use of their social, geolocation, or online data in AI training or decision-making.
• Purpose Limitation: Use of data for purposes other than originally intended—such as using social activism history to deny bail—violates the principle of purpose limitation enshrined in the DPDP Act.
• Lack of Safeguards: The Act does not mandate audit trails, data minimization, or human-in-the-loop decision-making, potentially leading to overreach.

III. Judicial Transparency and Right to Information

The Right to Information Act, 2005 mandates transparency and public access to state functioning. However, JusPredict’s algorithmic logic is shrouded under corporate secrecy claims. When a tool influencing liberty operates beyond the scope of public oversight, it directly contradicts the ideals of an open justice system.

If the judiciary relies on private AI tools without revealing their methodology, the public cannot evaluate whether justice is truly being served or merely predicted. This disempowers not only the accused but also their counsel, hindering effective legal defense.

IV. The Problem of Judicial Compulsion and Erosion of Discretion

AI tools like JusPredict can create a chilling effect on judicial discretion. Judges may feel constrained to conform to algorithmic recommendations to avoid institutional disapproval, appellate scrutiny, or reputational risk. This undermines the judiciary’s constitutional role as the final arbiter of facts and law.

The principle laid down in S.R. Bommai v. Union of India that judicial review must be based on real, verifiable grounds—not on speculative or external influences—is undermined when AI risk scores become decisive without independent judicial reasoning.

V. Need for Statutory Oversight and Ethical Auditing

Despite the transformative potential of AI, India currently lacks:

• A statutory framework for AI auditing in judicial contexts
• Accountability mechanisms for algorithmic errors or wrongful detentions
• Ethical review committees or ombudsman systems to assess biases
• Appeal mechanisms specifically for algorithm-influenced decisions

Internationally, jurisdictions such as the EU have emphasized Explainable AI (XAI) and Human-in-the-Loop governance—norms that remain absent in India’s framework.

Conclusion: Balancing Innovation with Constitutional Morality

The AIJDA 2023, while a bold experiment in modernizing the judicial system, must not come at the cost of constitutional values. Algorithmic tools like JusPredict may offer efficiency, but justice requires transparency, accountability, and fairness.

India must adopt a regulatory architecture that ensures:

• Mandatory human oversight of AI decisions
• Explainability and auditability of algorithmic tools
• Right to challenge algorithmic scores in judicial proceedings
• Data protection consistent with individual autonomy and consent