AI In Indian Courts

India’s judiciary is in a state of chronic overload. As of 2026, over 5.4 crore cases remain pending across courts, and litigants often wait years, sometimes decades, for a final outcome. The system is not broken in any dramatic sense rather it is worn down by heavy volume, and ordinary people bear the cost of that exhaustion mostly. It is against this background that the conversation around artificial intelligence in courts has acquired urgency.

That urgency, however, comes not only from the promise of efficiency but also from the risks of misuse. Courts in India have already encountered the problem of AI-generated hallucinations in legal proceedings. In December 2024, the Bengaluru Bench of the Income Tax Appellate Tribunal had to recall an order in the Buckeye Trust matter after it cited non-existent judgments generated through ChatGPT. Similar incidents followed in the Delhi High Court and the Bombay High Court, where fabricated citations and phantom paragraphs found their way into proceedings. These episodes exposed a basic but serious danger, AI can mimic the language and form of legal authority without any guarantee of accuracy.

It is in this context that, on 3 June 2026, the Supreme Court of India released the Draft Regulations for Use of Artificial Intelligence in Courts and invited public comments. The document is ambitious in both scale and intent. It attempts to build a governance framework for the use of AI in the judicial system, covering tasks such as legal research, translation, scheduling, and summarisation. That is a significant step by an institution, and it deserves careful attention, not only for what it enables, but also for the limits it tries to impose.

WHAT THE DRAFT GETS RIGHT

The strongest part of the Draft lies in its absolute prohibitions under Regulation 20. In particular, Regulation 20(1)(d), which prohibits AI from risk scoring which is the systemic way to assigning numbers to potential threats, bail eligibility assessments, and credibility determinations, is a sound and principled safeguard because these functions directly affect personal liberty and demand human judgment at the highest level.

The requirement of prior written approval before any AI system set out in Regulations 18 and 19(1), is another sensible safeguard, since it creates a checkpoint at which performance, reliability, and risk can be assessed before the technology affects actual litigants.

The Draft’s broader orientation also deserves credit. Regulation 16 creates a presumption in favour of responsible AI adoption, while Regulation 17 encourages innovation where it can improve access to justice and administrative efficiency. Taken together, these provisions show that the Draft is not trying to keep AI out of courts altogether; it is trying to regulate it before misuse becomes routine.

WHERE THE DRAFT FALLS SHORT

The Draft is thoughtful in many respects, but its biggest weakness lies in an assumption that does not match reality. It assumes that every court in India is ready to adopt AI on the same terms. It treats the judiciary as though the Supreme Court, a well-resourced High Court, and a district court in an under-equipped complex all operate under similar conditions. Matter of fact, they do not. The Department of Justice’s August 2024 empirical report showed that only 45% of judicial officers had electronic display facilities, nearly one-third of court complexes lacked video-conferencing capability, and 41% of support staff had no functional computer. A 2023 survey under the then Chief Justice Chandrachud, also found that 27% of judges did not own a digital device, while 10% of courts had no internet connection at all.

In that setting, a uniform regulatory framework can become more symbolic than practical. It may create the appearance of compliance without the infrastructure needed to make compliance real. A phased readiness threshold would address this problem more honestly. Before the obligations under Regulations 22 to 45 of institutional mechanism begin to operate, courts should first be assessed for basic digital capacity. Regulation should follow readiness, not replace it.

A second concern is that the Draft’s definitions are too broad for everyday use. It defines AI so widely, and “court process” so expansively, that even ordinary scheduling software or automated reminders may fall within the same regime as tools that actually shape judicial outcomes. That is too blunt an approach. A more workable framework would draw a line between AI that materially affects a decision or legal reasoning, and tools that only support routine administration. The former should be regulated closely; the latter should not be pulled into the same heavy architecture.

The verification safeguard in Regulation 8(3) also leaves room for concern. On paper, it is a sensible rule that AI outputs must be verified before use. But the proviso allows verification to be waived if reasons are recorded in writing, without setting any minimum standard for when that waiver is appropriate. That creates a gap large enough to weaken the safeguard itself. The risk is not theoretical. Indian courts Indiahave already seen fabricated citations and false authorities enter proceedings because someone relied on AI without checking the output. A verification rule is only as strong as the discipline behind it.

The Draft’s commitment to inclusion is also underdeveloped. Regulation 13 speaks of accessibility and fairness, but it does not explain how those goals will be achieved in courts where digital access remains uneven. For a large part of the population, especially in rural and economically weaker communities, technology can widen exclusion as easily as it can reduce delay. The existing e-Sewa Kendra network could provide a practical foundation here, but only if the Draft ties its promises to clear implementation duties.

Finally, there is a constitutional concern. Article 145 gives the Supreme Court power to frame rules for its own practice and procedure, but a framework that extends across High Courts, affects advocates, and shapes judicial administration nationwide may need stronger statutory backing. In that sense, the Draft is important, but still unfinished. It shows real promise, what it now needs is a closer fit with the realities of Indian courts on the ground.