AI-generated material can assist legal work, but it cannot displace the duty of lawyers and adjudicators to verify what is placed on record.
| Photo Credit:
bl-online Administrator
The Supreme Court may have overstretched the metaphor while it compared AI-hallucinated material to the release of “methyl isocyanate” in a recent matter where NCLT and NCLAT relied on AI-generated fictitious judicial precedents. But the judges were not mistaken in their assessment of the malaise.The Court’s observations were made in an insolvency dispute; it overturned NCLT and NCLAT rulings after discovering that they were based on non-existent, AI-hallucinated precedents. The judges were unequivocal in concluding that there has to be “zero tolerance” towards production, citation and use of AI-generated case laws without verification. While the Court characterised the actions of the lawyers who relied on non-existent case laws as “misconduct”, it stopped short of holding them in contempt or otherwise censuring them. Likewise, in setting aside the tribunal orders, it described the tribunals’ reliance on fabricated judgments merely as a “lapse”. The restraint is perhaps explained by the absence of a clear regulatory framework governing the use of generative AI in legal practice. The Court chose to address the larger institutional vacuum by directing the Bar Council of India to frame norms governing the use of AI by lawyers.Developments elsewhere suggest that such safeguards are in place. In the UK, the judgment in Ayinde v Haringey and Al-Haroun v Qatar makes it clear that publicly available generative AI tools based on large language models, including ChatGPT, are not reliable substitutes for legal research. Their output cannot be treated as authoritative unless independently verified. Lawyers who fail in that duty expose themselves to sanctions ranging from public criticism and adverse costs, to referrals to regulators, contempt proceedings and, in appropriate cases, criminal investigation. At the same time, regulators have sought to accommodate the technology within a supervised framework. In 2025, the Solicitors Regulation Authority authorised Garfield Law Ltd, to look into compliance systems in this regard. The US has adopted a strict approach. In Mata v Avianca, a New York district judge fined two lawyers $5,000 after they filed a brief containing fictitious authorities generated by ChatGPT. The court emphasised that while the use of AI is not inherently improper, lawyers cannot abdicate their duty to verify claims.Indian law provides for serious consequences where false material is placed before a court, including proceedings for contempt or perjury. The Bar Council should establish clear standards for the use of AI in legal practice. Equally, courts must point out that the use of AI does not dilute existing professional obligations. AI-generated material can assist legal work, but it cannot displace the duty of lawyers and adjudicators to verify what is placed on record. Where fabricated authorities are presented before a court, the issue is ultimately not one of technology but of accountability. The law should respond accordingly.Published on July 7, 2026
