“AI Can Assist, But It Cannot Decide”: Shardul Shroff at Harnessing AI to Transform India’s Judicial Ecosystem

Speaking at the India AI Impact Summit 2026 on 'Harnessing AI to Transform India’s Judicial Ecosystem', Dr. Shardul S. Shroff, Executive Chairman of Shardul Amarchand Mangaldas & Co., addressed the critical question of whether Artificial Intelligence can meaningfully reduce the burden on courts by strengthening dispute resolution mechanisms outside traditional adjudication.

Responding to a query on how AI tools can enhance efficiency, predictability and cost-effectiveness in arbitration and mediation, Shroff began by acknowledging the technological advantages. According to him, AI significantly improves “the speed of decision-making, the ability to collect data, to test data,” and to process large volumes of material in a compressed time frame.

He identified procedural functions where AI can be particularly effective: segregation of evidence, collection and management of documents, preparation of lists of dates, and due diligence on available data. These, he noted, are essentially procedural matters where machines can drastically reduce time consumption.

However, Shroff was unequivocal that AI must not cross into the realm of adjudicatory power.

He raised what he described as the “black box” issue: the opacity surrounding how AI systems arrive at outputs. The central concern, he said, is whether a judgment is ultimately delivered by mechanical means or by a human being. “If the decision is passed over to the machine, it would be bad law,” he cautioned, pointing out that no one has delegated adjudicatory authority to a machine.

In arbitration and mediation, parties consciously delegate authority to a named individual. “Has the delegation of authority been made to the person individually?” he asked, underscoring that such delegation “has not been made to the machine.” This distinction, he suggested, is fundamental. If machines begin influencing or determining outcomes, it could raise serious legal concerns, potentially even becoming a ground for setting aside an arbitral award.

Transparency and disclosure, according to Shroff, are central safeguards. AI-generated content must be traceable to the record, and there must be clarity about what data has been used. He warned against allowing systems unrestricted access to external material, which could introduce inaccuracies into the adjudicatory process.

Shroff also addressed the problem of AI “hallucinations,” describing situations where machines scrape data and generate incorrect or fabricated outputs. This can result in wrong parties being included, incorrect judgments being cited, or contextually inaccurate material entering proceedings. He observed that courts have encountered instances where lawyers relied on such faulty AI-generated material.

The solution, he emphasised, lies in human supervision. It should not be a “hands-free” kind of situation. AI use must not be “open-ended.” Lawyers remain professionally responsible for verifying outputs and ensuring accuracy.

Referring to evolving judicial approaches, Shroff noted that courts have begun articulating procedural principles governing AI usage. These include confining systems strictly to the record, preventing them from inventing or inferring legal opinions, ensuring that outputs are traceable to the source material, and maintaining transparent logs of interactions. Such safeguards, he suggested, are essential to preserving procedural integrity.

He further stressed that in dispute resolution, party autonomy remains paramount. If AI tools access external material beyond what parties have agreed upon, it could compromise fairness. The extent of AI usage must therefore be settled upfront, particularly in mediation and arbitration contexts.

In his concluding remarks, Shroff reiterated that while AI offers immense benefits in terms of efficiency and cost-effectiveness, it must remain an assistive tool. The power of decision-making, he maintained, cannot be transferred to a machine.

The discussion reflected a broader theme of the conference: technological transformation must operate within constitutional and procedural boundaries. AI may accelerate processes and enhance analytical capacity, but adjudicatory legitimacy, Shroff made clear, ultimately rests with the human decision-maker.

Event: India AI Impact Summit, 2026

Date: February 16, 2026