The moderator opened with a confession. Ms. Nivedita Chauhan, Founder and Managing Partner of NC Legal, told the room that a client had recently asked whether she used AI tools for legal research and drafting, and that she had hesitated to admit it, unsure how a lawyer using such technology would be perceived.
Two years ago, that hesitation was universal. At Parul Institute of Law‘s Law Conclave 2026, the panel she convened made clear that it has evaporated. The question is no longer whether Indian lawyers use artificial intelligence. It is where they may use it, and who is responsible when it fails.
The Debate Has Already Moved On!
Mr. Gauhar Mirza, Partner at Saraf and Partners, set the terms. Artificial intelligence, he argued, is not a future consideration. Lawyers, clients, arbitrators, and judges are already using AI tools in their daily work. The profession should therefore stop asking whether AI should be used and start learning to use it responsibly.
He pointed to draft guidelines under consideration by the Supreme Court, which centre on human control, responsibility, fairness, privacy, cybersecurity, and data protection. AI can assist with research, document review, and case preparation. The final decision, and the responsibility for it, must always rest with a human lawyer.
Mr. Harry Chawla, Managing Partner of Luthra and Luthra Law Offices, described how fast the ground has shifted. When clients first began asking law firms about AI, many lawyers were uncomfortable admitting they used it. Today, firms openly acknowledge working with licensed AI tools, treating them as technology that supports rather than substitutes for legal work.
PU Law Conclave 2026 – Bhumika Induliya’s experience on LinkedIn!
The Slave and the Master
Mr. Vivek Kohli, Senior Advocate and former Advocate General of Sikkim, offered the principle that anchored the panel. AI is a useful tool, but it must never control legal decision-making.
“As a slave and not as a master” – Mr. Vivek Kohli, Senior Advocate, on how AI should serve the legal process
He then did something more useful than assert a principle. He divided the work of a lawyer into three categories and assessed AI’s fitness for each.
- Administrative and routine work: arranging documents, connecting information, managing large volumes of data, and presenting it simply. Tasks once handled by junior lawyers with zero to five years of experience can now be done faster and better with AI.
- Analytical and supporting work: surfacing relevant material, offering suggestions, and helping a lawyer navigate difficult legal texts. Human thinking remains essential. Understanding facts, weighing evidence, and reaching legal conclusions stay with the lawyer.
- Decision-making and adjudication: the most sensitive category, and the one where Kohli was firmest. AI cannot replace human decisions in arbitration or litigation. It can find errors and organise evidence, but it cannot grasp human emotion, empathy, or circumstance, and justice requires all three.
His caution came from practice rather than theory. He described using AI in large oil and gas arbitrations involving thousands of documents, where it helped him absorb technical material and manage information at scale. Every output, he noted, was verified by experts and clients before it entered a legal proceeding.
“Green zone: research. Yellow zone: summaries, with human review. Red zone: final pleadings. Never unchecked.”
The Three-Zone Framework
Mr. Ravi Singhania, Managing Partner of Singhania and Partners LLP, translated the principle into a rule a firm could actually adopt. Before AI enters serious judicial work, he argued, courts must define clearly where it may and may not be used, and he proposed three zones.
- Green zone: basic legal research and straightforward legal analysis, where AI can be used freely.
- Yellow zone: preparing case summaries, building timelines, and managing documents, where AI may be used but human verification is mandatory.
- Red zone: filing final pleadings or relying on AI-generated legal submissions without proper checking. These decisions belong to trained legal professionals, always.
The framework is valuable precisely because it is boring. It does not moralise about technology. It tells a junior associate what they may do on a Tuesday afternoon, which is what a professional standard is for.
What AI Cannot Fix: The Execution Problem
The moderator raised the practical failure of Indian arbitration. Although arbitration exists to resolve disputes outside the courts, parties must still approach civil courts to execute an arbitral award, and that process is routinely slow. Could execution be brought inside the ADR system, and could AI accelerate it?
Mr. Ravi Singhania was blunt. Delay in execution is among the biggest problems in Indian arbitration, but expecting AI to solve it soon is unrealistic because the judicial system is only beginning to adopt these tools and has not yet written the rules governing their use.
Mr. Vivek Kohli reframed the problem entirely. Instead of accelerating the end of the process, he argued, fix the beginning. The deeper flaw in Indian dispute resolution is that reliable information about a party’s assets is not available before proceedings start. Arbitration can run for years, during which a losing party may move or dissipate assets, so the winner holds an award and has no means of enforcing it.
AI, he suggested, should be used as a preventive tool: searching public databases to establish the assets and financial position of parties at the outset, strengthening financial disclosure before a dispute begins rather than chasing assets after it ends. Mr. Mirza agreed, noting that AI can trace assets during enforcement while insisting again that it supports human experts rather than replaces them.
The Award That Was Set Aside, and Who Was to Blame
The sharpest moment came from the audience. A participant noted that an arbitral award by a respected Indian arbitrator had been set aside by the Singapore courts because passages had been copied from earlier judgments, a problem that predates AI entirely. Why, he asked, has Indian arbitration not attained the standing of Singapore or London?
Mr. Kohli answered first and disclosed that he had personally been involved in that case before the Singapore courts. The award was not set aside because the legal reasoning was wrong, he explained, but because portions had been reproduced from earlier judgments without independent analysis. Had proper checking tools been used, the defect could have been caught.
His conclusion was the panel’s central ethical claim. AI is not to blame for such failures. Responsibility always rests with the person using the technology. Whether material is drawn from an AI system, an earlier judgment, or any other source, judges, arbitrators, and lawyers must verify it and apply their own legal reasoning before deciding.
Mr. Singhania pushed back on the premise of the question. The belief that arbitration in Singapore and the United Kingdom is flawless while India’s is weak, he argued, survives because Indian failures attract attention that foreign ones do not. Awards are challenged and set aside elsewhere too, and Indian courts handle an enormous caseload with limited resources. Mr. Chawla agreed that no arbitration system anywhere is free of error, citing foreign awards annulled over arbitrator conflicts of interest, and advised young lawyers to spend their energy improving India’s system rather than comparing it unfavourably with others.
Mr. Mirza located the real weakness precisely. The quality of Indian arbitral awards is generally high, and Indian arbitrators are experienced. The problem is not the quality of decisions but the time taken to complete cases and enforce awards. That, he argued, is where AI can genuinely help.
The Structural Fix Nobody Wants to Talk About
Before the session closed, Mr. Chawla named a problem no algorithm can solve. India’s arbitration culture depends too heavily on ad hoc arbitration. Greater use of institutional arbitration would improve consistency, monitoring, and the quality of proceedings. It is a structural reform, not a technological one, and it sits alongside the constitutional questions taken up in the conclave’s panel on surveillance and liberty as evidence that the digital era’s hardest legal problems are rarely solved by software. So if you wish to learn the authenticity of the legal domain with such leading dignitaries of India, enrol on BA LLB, BBA LLB, and LLM programmes at Parul Institute of Law, where you’ll gain the legal knowledge, courtroom exposure, and practical skills to thrive in a digital era!
Frequently Asked Questions
Can lawyers use AI for legal work in India?
Yes, and many already do. At Parul University’s Law Conclave 2026, India’s senior arbitration practitioners confirmed that lawyers, clients and judges are using AI tools and they’re drafting them in sync with the guidelines of the Supreme Court, but the final decision is always executed by a human lawyer!
Where is it safe to use AI in legal practice?
Mr. Ravi Singhania proposed a three-zone framework at the conclave. A green zone covers basic legal research and simple analysis, where AI can be used freely. A yellow zone covers case summaries, timelines, and document management, where human verification is mandatory. A red zone covers final pleadings and legal submissions, which must never rely on unchecked AI output.
Can AI replace arbitrators or judges?
No, according to the panel. Senior Advocate Vivek Kohli argued that AI can organise evidence and identify errors, but cannot understand human emotion, empathy, or circumstance, and that adjudication requires all three. AI, in his phrase, should function as a slave and not as a master.
Why are arbitral awards in India slow to enforce?
Because parties must approach civil courts to execute an award, a process that faces long delays. Panellists argued that a deeper cause is the absence of reliable information about parties' assets before proceedings begin, allowing a losing party to move assets during years of arbitration. AI was proposed as a preventive tool for asset discovery at the outset.
