J
In Conversation With
Justice (Retd.) P.K. Vishwanathan
Former Judge, Supreme Court of India
Interview conducted by
Staff Reporter
· May 8, 2026
**TestLaw:** Justice Vishwanathan, you recently spoke at the SILF conference on AI in courts. What is your central concern?
**Justice Vishwanathan:** My central concern is that we are already seeing AI-generated documents, AI-summarised case law and even AI-drafted plaints being submitted in courts — without any disclosure. If a lawyer submits a brief that was substantially written by an AI and that AI has hallucinated a citation, who is responsible? The court, the lawyer, the AI company? We have no framework for this.
**TestLaw:** Should the Supreme Court issue guidelines?
**Justice Vishwanathan:** I would prefer the Bar Council of India and the Bar Associations to act first. Courts are reactive institutions — they rule on disputes that come before them. The Bar is a self-governing profession. It should govern itself. The danger is that if the bar does not act, a litigant will be prejudiced by an AI error, they will complain, and then the court will be forced to frame rules in the heat of a specific controversy. That is the worst way to make policy.
**TestLaw:** What specific rules would you recommend?
**Justice Vishwanathan:** Three things at minimum. First, mandatory disclosure when AI has been used to draft or research any document submitted to a court. Second, the certifying advocate must personally verify every citation regardless of how it was generated. Third, the Bar Council should maintain a list of approved legal AI tools that have been tested for accuracy with Indian case law. Right now, most of these tools have been trained primarily on US and UK data and they are unreliable on Indian statutory interpretation.
**TestLaw:** How receptive is the judiciary to this?
**Justice Vishwanathan:** More than you might think. I have spoken to several sitting judges who are themselves using AI tools to research background to cases — not for judgment-writing, they are careful about that — but for understanding technical subjects. One judge told me he used an AI tool to understand the technical specifications in a patent dispute and found it very useful. But judges also know the risks. I think there is genuine appetite for a sensible regulatory framework. The ball is in the Bar's court.
**TestLaw:** What about AI-generated evidence — deepfakes, synthetic audio?
**Justice Vishwanathan:** This is the most urgent frontier. We had no statutory framework for electronic evidence authentication that keeps pace with the sophistication of forgery today. The Indian Evidence Act amendments help somewhat, but the courts need forensic infrastructure — certified laboratories that can examine AI-generated content. Without that, we will have judges ruling on deepfakes without the technical tools to verify them. That is a recipe for injustice.
**Justice Vishwanathan:** My central concern is that we are already seeing AI-generated documents, AI-summarised case law and even AI-drafted plaints being submitted in courts — without any disclosure. If a lawyer submits a brief that was substantially written by an AI and that AI has hallucinated a citation, who is responsible? The court, the lawyer, the AI company? We have no framework for this.
**TestLaw:** Should the Supreme Court issue guidelines?
**Justice Vishwanathan:** I would prefer the Bar Council of India and the Bar Associations to act first. Courts are reactive institutions — they rule on disputes that come before them. The Bar is a self-governing profession. It should govern itself. The danger is that if the bar does not act, a litigant will be prejudiced by an AI error, they will complain, and then the court will be forced to frame rules in the heat of a specific controversy. That is the worst way to make policy.
**TestLaw:** What specific rules would you recommend?
**Justice Vishwanathan:** Three things at minimum. First, mandatory disclosure when AI has been used to draft or research any document submitted to a court. Second, the certifying advocate must personally verify every citation regardless of how it was generated. Third, the Bar Council should maintain a list of approved legal AI tools that have been tested for accuracy with Indian case law. Right now, most of these tools have been trained primarily on US and UK data and they are unreliable on Indian statutory interpretation.
**TestLaw:** How receptive is the judiciary to this?
**Justice Vishwanathan:** More than you might think. I have spoken to several sitting judges who are themselves using AI tools to research background to cases — not for judgment-writing, they are careful about that — but for understanding technical subjects. One judge told me he used an AI tool to understand the technical specifications in a patent dispute and found it very useful. But judges also know the risks. I think there is genuine appetite for a sensible regulatory framework. The ball is in the Bar's court.
**TestLaw:** What about AI-generated evidence — deepfakes, synthetic audio?
**Justice Vishwanathan:** This is the most urgent frontier. We had no statutory framework for electronic evidence authentication that keeps pace with the sophistication of forgery today. The Indian Evidence Act amendments help somewhat, but the courts need forensic infrastructure — certified laboratories that can examine AI-generated content. Without that, we will have judges ruling on deepfakes without the technical tools to verify them. That is a recipe for injustice.