In a criminal justice system operating under immense institutional pressure, Human-in-the-Loop AI has to be approached with an open mind
JAASIR ASHRAF MIR
The Supreme Court’s decision, while granting bail to the accused in Syed Iftikhar Andrabi v. National Investigation Agency (2026), has brought in again in discussions a fundamental question in Indian criminal law pertaining to the role of courts in promoting consistency in the exercise of judicial discretion while preserving judicial independence, constitutional guarantees, and the flexibility necessary to achieve justice in individual cases. The immediate context of this question arises prominently in bail adjudication under the Unlawful Activities (Prevention) Act, 1967, particularly Section 43D(5) of UAPA; however, it extends across criminal procedure more broadly.
That the Supreme Court’s Draft Regulations for the Use of Artificial Intelligence (AI) in Courts, 2026, has further made the issue expedient.
In this article, I make an exploration, whether a carefully regulated Human-in-the-Loop AI framework can assist judicial decision-making. Further, it argues that whether AI must never replace judicial reasoning or determine outcomes, even if it may help reduce arbitrariness, promote greater consistency, and support a more structured balancing of competing interests affecting personal liberty, thereby strengthening the rule of law.
Judicial Discretion, Personal Liberty, and the Quest for Consistency in Bail Adjudication
Bail adjudication under the UAPA is not often a routine procedural exercise. It calls upon the courts to balance personal liberty, national security, evidentiary thresholds, and constitutional rights. Central to such bail matters lies a continuing tension wherein courts must safeguard the individual’s right to personal liberty under Article 21 of the Constitution while simultaneously addressing concerns relating to national security, public safety, the integrity of criminal investigations, and the effective administration of justice. Bail jurisprudence, therefore, operates within a framework shaped ordinarily by competing constitutional considerations and demands a harmonious approach.
In the final analysis of such determination, Judicial discretion remains central to the exercise. Criminal cases differ in their facts, evidentiary records, procedural histories, and social contexts. Any attempt to eliminate discretion would equally risk undermining the ethos of justice itself. However, at the same time, discretion presents institutional challenges equally. In situations where similarly situated individuals approach different courts or even different benches of the same court, outcomes may vary considerably. The concern, therefore, is not the existence of discretion but the manner in which discretion is exercised and applied. A legal system committed to equality before the law and equal protection of the laws benefits from coherence, predictability, and reasoned consistency in judicial decision-making.
In the Nature of the Judicial Process, Justice Benjamin Cardozo offers important insights into judicial decision-making. He explains that judicial decision-making is not a mechanical exercise involving the simple application of pre-existing rules. Rather, in cases where the law is incomplete or unsettled, judges inevitably draw upon precedent, logic, history, custom, and considerations of social utility. At the same time, Cardozo recognised that in straightforward cases, where clear statutory language or binding precedent exists, the judicial role is largely confined to applying the law to the facts before the court.
The concern of discretionary power or, for that matter, the debate regarding it, is not new; it is as old as the laws. However, Andrabi’s case, when juxtaposed with the Supreme Court’s AI draft regulation, provide us an opportunity to kick start the discussion regarding the exercise of discretionary power under the UAPA. More so, considering the inconsistency flagged by the Supreme Court in the Andrabi judgment.
The Supreme Court’s decision in Union of India v. K.A. Najeeb (2021) marked a significant constitutional intervention in UAPA bail Jurisprudence. In Najeeb, the Court laid down the principle that constitutional courts are not rendered powerless merely because Parliament has enacted restrictive bail provisions. The Court held that prolonged incarceration and a delayed trial may itself amount to a violation of Article 21. This situation leads to the continuing tension between the constitutional guarantee of personal liberty under Article 21 and statutory provisions that impose stringent restrictions on the grant of bail. In appropriate cases, therefore, continued detention may raise constitutional concerns under Article 21 even under a special statute containing a stringent embargo on bail.
Following Najeeb, courts have been required to navigate the relationship between two distinct standards. The first is the statutory threshold under Section 43D(5) under UAPA, which requires courts to determine whether there are reasonable grounds for believing that the accusation is prima facie true. The second is the constitutional obligation to protect personal liberty and guard against indefinite incarceration. Subsequent decisions, including Gurwinder Singh v. State of Punjab (2024) and Gulfisha Fatima v. State (Govt. of NCT of Delhi) (2026), illustrate how courts have approached this tension differently in practice, particularly where the statutory embargo continues to weigh heavily despite significant delays in trial. The jurisprudence emanating from the Najeeb, Gurwinder Singh, and Gulfisha has therefore produced an inconsistent jurisprudence leading to confusion, which doesn’t abode well for the rule of law.
Following Najeeb, the SC in Andrabi appears to continue the constitutional balancing exercise. The Supreme Court, while granting bail to Andrabi, reaffirmed that the principles articulated in Najeeb in 2021 remain relevant where incarceration has become prolonged and the prospects of trial appear remote. Arguably, the judgment may be read as an effort to preserve the primacy of liberty while remaining attentive to the statutory framework of the UAPA. Whether the precedent set now reflects continuity or a context-specific refinement is a matter on which reasonable views may differ.
Across India’s criminal justice system, the underlying challenge is broader than any single statute. Judges are routinely required to decide bail applications, remand matters, anticipatory bail petitions, and other pre-trial issues under substantial institutional constraints. Heavy caseloads, voluminous records, limited hearing time, and an ever-expanding body of precedent create conditions in which variations in reasoning and outcomes may be difficult to avoid. If consistency is to be improved without reducing the quality of the verdict into a mechanical exercise, the question to be considered now is how assistive technology can support judicial reasoning without displacing it.
Human-in-the-Loop AI and the Supreme Court’s Draft Regulatory Framework
Although the Supreme Court’s draft Regulations for Use of Artificial Intelligence (AI) in Courts, 2026, adopts a carefully human-centric framework and reflects a clear preference for judicial primacy. However, there is still limited empirical evidence in India to show whether Human-in-the-Loop artificial intelligence can meaningfully improve consistency, promote greater uniformity in the proportionality-based balancing of competing interests, reduce cognitive burdens on judges, or otherwise enhance the quality of judicial decision-making. These questions therefore warrant being addressed with an open mind and thereof requires closer and more rigorous study, particularly in light of the safeguards, exclusions, and institutional concerns reflected in the draft regulations.
Constitutional Safeguards, Judicial Primacy, and the Future of Criminal Justice
Indeed, the safeguards accompanying any technological intervention in the justice delivery system are important. However, if the use of technology and use of judicial primacy is used in a way which will aid in protecting constitutional values, consistency, transparency, explainability, auditability, data integrity, accountability, and strengthen the rule of law, it can revolutionise the whole justice system. The draft regulations is rightly a significant development and providing safeguards regarding the use and against automation bias, which is also essential, since judges may, in practice, attach undue weight to outputs that appear objective or technologically sophisticated. Since the legal processes become increasingly digital, interaction between technology and adjudication may become unavoidable.
The more important question, therefore, is not whether technology should enter the courtroom. In many respects, it already has, and we experience it. The real issue is how it should be structured so that it supports constitutional values rather than weakening them.
The recent trajectory of bail jurisprudence quoted above under the UAPA helps explain why this inquiry matters. In a legal system where liberty often depends upon the interpretation of precedent, and where similar factual situations can sometimes produce different outcomes, consistency is not merely an administrative preference. It is closely connected to equality before the law, procedural fairness, and the rule of law itself.
Human-in-the-Loop AI may or may not eventually provide part of the answer. Even so, the scale, complexity, and institutional demands of India’s criminal justice system make it a question worthy of serious consideration. The issue should perhaps be viewed not as a technological solution to a legal problem, but as a constitutional and institutional inquiry into how courts might preserve coherence, consistency, transparency, proportionality and fairness in the administration of justice while strengthening the rule of law.
The Supreme Court’s draft Regulations for Use of Artificial Intelligence (AI) in Courts, 2026, is therefore a significant development. The article lays more emphasis on all stakeholder participation in the consultation process, as it is likely to be crucial if the eventual framework is to support the rule of law while preserving judicial reasoning, judicial independence, and the human responsibility that remains central to adjudication.
Draft Regulation on AI as a way forward and Preserving Judicial Primacy
The active participation of experts and all stakeholders is important as to make the final regulations on AI useful, which will result in extracting maximum benefits from technological uses without compromising on a human-centric framework and reflecting a clear preference for judicial primacy. Today, the challenge is not one between judges and machines. Rather, it is how to reduce arbitrariness, minimise inconsistency, and ensure that the proportionality-based balancing of competing interests affecting personal liberty is undertaken in a more structured, transparent, and principled manner that strengthens the rule of law.
In a criminal justice system operating under immense institutional pressure, Human-in-the-Loop AI has to be approached with an open mind and may offer one possible means of advancing these objectives while preserving judicial primacy. Whether it can do so effectively, however, remains a question that warrants sustained empirical and normative inquiry, one that Indian criminal jurisprudence can no longer afford to ignore.
(The Author is LLM, SOAS as a Chevening Scholar (2023–24). Senior Prosecuting Officer, Government of Jammu and Kashmir)
