AI in judiciary- Has the time come?

The debate over whether artificial intelligence (AI) should enter Indian courts is no longer relevant—it is already happening. AI tools are actively being deployed across the judicial system, from translation engines to real-time transcription services. The more pressing question now is how the Indian legal system should respond to this reality. Responses so far have been uneven, varying from bureaucratic caution and outright paranoia to overly casual reliance on these technologies. A common thread running through most approaches is a deep-seated anxiety: the fear that AI will introduce errors, inaccuracies, or "hallucinations" into the sacred work of judging cases. 

This fear, as per a section of legal luminaries is largely disproportionate. The Indian judiciary, they say, has never pretended to be infallible. Instead, it has always depended on built-in mechanisms of accountability, self-correction, appeals, reviews, and scrutiny to address human fallibility. These safeguards should not suddenly fail simply because the tools assisting judges change from manual research to AI-assisted processes. As per them, modernization through technology, including AI, should be embraced structurally rather than treated as a conditional experiment awaiting perfect reliability.

A senior advocate opined that several initiatives already demonstrate AI's productive role in the judiciary. He also added that the Supreme Court's translation engine has rendered over 36,000 judgments into 19 Indian languages, making justice more accessible across linguistic barriers. Tools like TERES provide real-time AI transcription for Constitution Bench hearings, while the Kerala High Court employs a transcription system developed by Adala.ai for witness depositions in all subordinate courts. The government's e-Courts Phase III project allocates around Rs 7,000 crores for judicial modernization, with a portion—approximately Rs 50 crores—dedicated to AI and emerging technologies. These efforts highlight that AI is not merely experimental but increasingly integral to improving efficiency and reach.

Yet, institutional responses reveal a persistent "fear of the machine." The Supreme Court recently admonished lawyers for citing hallucinated cases generated by AI. The Kerala High Court's policy on AI use in the district judiciary correctly emphasizes that ultimate responsibility for orders rests with the judge—a crucial principle. However, it imposes heavy procedural burdens, such as detailed audits, mandatory ledgers recording every AI use, and reporting of errors to the principal district judge. The Supreme Court's White Paper on AI in the Judiciary, released in November 2025, offers a more balanced view: AI should serve the ends of justice without defining them, with judges remaining the ultimate decision-makers and human oversight as foundational. Even here, though, there is a tendency to add layers of procedural requirements.

This instinct to regulate through process-heavy controls stems from a concern that AI could corrupt judicial proceedings in unprecedented ways. A recent example from the Manipur High Court illustrates the point: in a case involving village defense forces, the court resorted to ChatGPT for research on the body's functions and limits—raising questions about why a constitutional court would rely on a free consumer chatbot rather than established resources. Such incidents fuel calls for tighter rules. But the better approach is not to pile on onerous procedures but to strengthen existing institutional capacities focused on outcomes rather than micromanaging processes.

A judge of the Madras High Court has taken a forward-thinking stance. In a recent order (around early 2026), he carefully examined legal tech products and permitted the use of "Superlaw Courts" in an arbitration matter for record-keeping and document management. He stressed that legal judgment must remain with human decision-makers. While his reasoning analogized AI functions to human workflows (e.g., comparing retrieval-augmented generation to preparing case extracts), this may unnecessarily concede that AI must justify itself by mimicking humans. A stronger frame asks simply whether the officer of the court can stand behind the final work product.

If Indian courts truly wish to modernize, they must move beyond the fear of the machine and trust the institution behind it. We choose the law and the fallible humans who administer it. Precisely because humans err, the system incorporates appeals, reviews, and accountability. No tool—however advanced—should undermine this fundamental compact. By focusing on responsibility and outcomes rather than procedural smothering, the judiciary can harness AI's benefits while preserving the human core of justice.