Supreme Court Justice Vikram Nath on Saturday cautioned that artificial intelligence (AI) can only supplement, not substitute, human intelligence in arbitration and adjudication.

“Technology can assist, but it cannot take over the role of human judgment...justice is, at its core, human. It requires what I call a heartbeat,” Justice Nath remarked, underscoring that machines may review documents and highlight patterns but lack fairness, empathy and conscience.

He was chairing a session on Arbitration 2.0: Integration of Artificial Intelligence and other technologies to enhance the efficiency of arbitration hosted by the Supreme Court of India, the High Court of Delhi and the Delhi International Arbitration Centre (DIAC) as a part of Delhi Arbitration Weekend 2025 (DAW).

The panel also included Madras High Court Justice N Anand Venkatesh, Solicitor General Tushar Mehta, independent arbitrator Mark Dempsey SC and Barrister at Brick Court Chambers Tim Lord KC.

Justice Nath placed the debate in historical perspective, recalling that arbitration itself was conceived as a response to the “slow grind” of traditional litigation. Over the years, the system moved from paper bundles and lengthy correspondence to scanned filings and online hearings, he said, adding,

“If arbitration was designed to save time, then AI has the potential to redefine what time even means in dispute resolution."

He drew attention to international practices such as the Chartered Institute of Arbitrators’ guidelines, which require disclosure whenever arbitrators use AI tools and the Singapore International Arbitration Centre’s Gateway Platform, aimed at digitising case management. At home, he noted, the Indian Supreme Court has already piloted AI in its own functioning — real-time transcription during Constitution Bench hearings, translation of judgments into regional languages and the Supreme Court Portal Assistance for Court Efficiency (SUPACE) to aid legal research.

“These examples show us how technology can strengthen our processes. But technology must operate within limits. It can make our system sharper and faster, but the conscience of justice — fairness, empathy and values — remains exclusively human," he noted.

Justice Nath also warned that misuse of AI could erode confidence in arbitration. He referred to a case in California where a party challenged an arbitral award on the ground that the arbitrator may have relied on AI to draft parts of the decision.

“The debate there was not about technology itself, but about accuracy, transparency and the integrity of the arbitral process,” he observed.

Justice Venkatesh said that arbitration had already been reshaped by digital databases, annotation tools and transcription services. But AI marked a leap from passive support to active participation.

“Artificial intelligence will revolutionise arbitration. It will not just execute commands; it will anticipate, analyse and sometimes even decide,” he observed.

He explained how AI could prepare chronologies, identify evidentiary gaps and test arguments functions that could transform efficiency. However, he cautioned that the shift raised fundamental questions:

“From being a tool or a passive supporter, AI is going to become an active participant, which will impact the core arbitral functions itself.”

SG Mehta warned against the risks of over-reliance on AI. He cited instances of AI producing fabricated judgments and case law, and flagged the danger of algorithmic bias.

“Hallucination of course is one problem, but bias is emerging as the second most scary problem. It’s fast going to be the first,” he said.

Lord KC discussed the rise of blockchain and smart contracts in arbitration, noting both their efficiency and the confidentiality risks they posed.

“It is important for every panel to rigorously endorse and police confidentiality of the proceedings,” Lord said, underlining that confidentiality remains central to arbitral integrity even in an age of technological disruption.

Dempsey SC warned tribunals against uncritical reliance on machine-generated outputs. He said,

"Because unless the parties are required to verify those submissions, it places an intolerable burden on the tribunal and it would be counterproductive to the efficiencies of the arbitral process if tribunals receive submissions created by AI knowing that they could be full of hallucinations and false records.”

At the conclusion of the panel discussion, SG Mehta, on a lighter note, recounted a story about how authors and judges are sometimes believed not to have written or even read their own works. He recalled how a popular novelist in America was rumoured to have his books ghostwritten, and drew a parallel with the judiciary.

“One of the judges who retired a few years ago had written a very brilliant judgment. A colleague asked me how I found it. I said I would not accuse him of having written it. Then he asked whether the judge may have at least read it before signing. I said no, he would probably read it when in the next case before him we started relying on that judgment,” Mehta quipped, drawing laughter from the audience.