I. Introduction
The legal profession has historically been slow to change, anchored in precedent, procedure, and a professional culture that prizes caution over disruption. Yet the emergence of generative artificial intelligence (AI) which is capable of drafting pleadings, summarizing case law, and even predicting litigation outcomes, it has compressed decades of technological adaptation into a few short years. Tools such as ChatGPT, CoCounsel, Harvey, and India’s own SUPACE (Supreme Court Portal for Assistance in Court’s Efficiency) are no longer peripheral curiosities; they are becoming embedded in everyday legal practice, from large corporate firms to solo practitioners. This raises a foundational question of professional responsibility: what does the ancient duty of competence mean when a lawyer’s tools can think, draft, and occasionally fabricate on their own?
This article examines how AI is reshaping the ethical obligations of lawyers, focusing on the duty of competence, confidentiality, and candour to the court. It draws on statutory frameworks (the ABA Model Rules of Professional Conduct, the Bar Council of India Rules under the Advocates Act, 1961, and emerging data protection legislation), judicial precedents from courts grappling with AI misuse, and scholarly commentary, to argue that existing ethical frameworks are adaptable to AI but require urgent regulatory clarification specific to the Indian context.
II. The Changing Landscape: AI in Legal Practice
Legal scholars have long anticipated the automation of legal work. Richard Susskind’s influential thesis in Tomorrow’s Lawyers predicted that legal services would be “commoditized” as routine tasks such as, document review, due diligence, legal research, became susceptible to technological substitution. More recent empirical work by Dana Remus and Frank Levy, in Can Robots Be Lawyers?, offered a more measured view: AI could automate discrete tasks within legal work, but the bundle of skills that constitutes lawyering which includes judgment, negotiation, empathy, and courtroom advocacy, all these would remain resistant to full automation for the foreseeable future.
Generative AI, however, differs qualitatively from earlier legal technology such as e-discovery software or predictive coding. Large language models generate fluent, persuasive prose that mimics legal reasoning without necessarily being grounded in verified legal authority. Harry Surden has described this as the central paradox of AI in law: the technology is exceptionally good at producing text that looks like sound legal analysis, while being unreliable at guaranteeing that the analysis is actually correct. This paradox is precisely what has generated the most significant ethical controversies to date, not AI replacing lawyers, but AI misleading them.
III. The Duty of Competence: Statutory and Regulatory Foundations
A. The American Framework
The clearest statement of a lawyer’s competence obligation in respect of technology comes from the United States. Rule 1.1 of the ABA Model Rules of Professional Conduct requires a lawyer to provide “competent representation,” defined as requiring “the legal knowledge, skill, thoroughness and preparation reasonably necessary. In 2012, the ABA amended Comment 8 to Rule 1.1 to state explicitly that competence “includes keeping abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology.” As of 2024, a majority of U.S. state bars have adopted some version of this “technology competence” duty, and several, including California, New York, and Florida, they have issued specific ethics opinions extending it to generative AI, requiring lawyers to understand an AI tool’s capabilities and limitations before relying on its output.
B. The Indian Framework
India’s regulatory architecture is comparatively underdeveloped in this respect. The Advocates Act, 1961, and the Bar Council of India Rules made thereunder impose general duties of diligence, competence, and candour to the court under Chapter II, Part VI of the Rules, but contain no reference to technology or AI. Rule 11, for instance, obliges an advocate not to act “in a manner unbecoming” of a counsel, while the general duty to the client under Section I requires the advocate to “fearlessly uphold the interests of his client” through “fair and honourable means”, a formulation broad enough to encompass diligence in verifying AI-generated material, but not specific enough to give practitioners clear guidance.
This regulatory silence is significant given the pace of judicial and institutional adoption of AI in India. The Supreme Court’s e-Committee has actively promoted AI tools such as SUPACE for judicial research assistance, and the SUVAS translation tool for multilingual judgments. Scholarly commentary, including work published in the NUJS Law Review and by NITI Aayog’s National Strategy for Artificial Intelligence, has called for the Bar Council of India to issue specific guidance on AI use, akin to the U.S. state bar opinions, but no binding rule has yet been promulgated.
IV. Judicial Precedents: Cautionary Tales of AI Misuse
The most consequential development in this area has come not from regulators but from courts sanctioning lawyers for the careless use of AI, cases that now function as de facto guidance on the boundaries of competent practice.
A. Mata v. Avianca, Inc.
The paradigmatic case is Mata v. Avianca, Inc., decided by the United States District Court for the Southern District of New York in 2023. Attorneys representing the plaintiff used ChatGPT to research and draft a brief in opposition to a motion to dismiss. The chatbot fabricated several judicial decisions complete with plausible-sounding case names, docket numbers, and quotations, none of which existed. The attorneys did not independently verify the citations before filing. Judge P. Kevin Castel sanctioned the lawyers, holding that “technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance,” but that lawyers “abandon their responsibilities when they submit papers without confirming their contents are accurate.” The case has since become the leading authority for the proposition that AI-assisted drafting does not dilute, but rather heightens, the duty of independent verification.
B. Park v. Kim
The Second Circuit Court of Appeals confronted a similar issue in Park v. Kim (2024), where counsel cited a non-existent case in an appellate brief generated with AI assistance. The court referred the matter to its Grievance Panel, observing that the duty of candour under Federal Rule of Civil Procedure 11 applies with undiminished force regardless of the tool used to prepare a filing, and used the occasion to caution the broader bar that “the duty to check … does not divide by task.”
C. The Indian Position: Jaswinder Singh v. State of Punjab
India’s engagement with this issue, while less punitive, has been notable. In Jaswinder Singh v. State of Punjab, a bail matter before the Punjab and Haryana High Court in March 2023, Justice Anoop Chitkara referred to a response generated by ChatGPT on the general approach to bail in cases involving cruelty, explicitly clarifying in the order that the AI-generated content was cited “only to have a broader perspective” and was “not an expression of opinion on the merits of the case,” and was not to be treated as a judicial finding. The order attracted considerable academic criticism, with commentators in the Indian Journal of Law and Technology arguing that even limited, clearly-labelled reliance on an unverifiable, non-transparent AI system risks eroding the reasoned, precedent-based character of judicial decision-making that Article 141 of the Constitution presupposes. The episode illustrates that the competence problem is not confined to the Bar; it extends to the Bench, and Indian regulators have yet to articulate a framework for either.
D. Algorithmic Bias and Due Process: State v. Loomis
A separate but related line of concern involves the use of AI in judicial decision-making itself, particularly risk-assessment tools. In State v. Loomis, the Wisconsin Supreme Court upheld the use of the COMPAS algorithm in sentencing, despite the defendant’s argument that the proprietary, “black box” nature of the tool violated his due process right to be sentenced on individualized, transparent grounds. The court permitted its use subject to cautionary jury/bench instructions about the tool’s limitations. The case remains contested precedent, and scholars such as Cathy O’Neil have argued that opaque algorithmic tools risk entrenching bias at scale precisely because they wear the appearance of mathematical objectivity. For the Indian context, where Article 21’s guarantee of a fair procedure “established by law” was substantially expanded in K.S. Puttaswamy v. Union of India to encompass informational privacy and protection against arbitrary algorithmic action, Loomis-type concerns are likely to surface as India’s judicial AI tools mature.
V. Confidentiality, Privacy, and the Duty of Loyalty
A further ethical dimension concerns the confidentiality of client information fed into AI systems. Many generative AI tools retain user inputs for model training unless expressly configured otherwise, raising the risk that privileged client data could be exposed or, in principle, surfaced in a future output to an unrelated user. Rule 1.6 of the ABA Model Rules imposes a strict duty of confidentiality, and several U.S. ethics opinions have now clarified that inputting client information into a public AI tool without adequate safeguards may itself constitute a breach.
India’s position is again comparatively unsettled. The Digital Personal Data Protection Act, 2023 imposes obligations on “data fiduciaries” to secure personal data and limit its processing to specified purposes, which would apply to a law firm’s use of client data in AI tools, but the Act does not address privilege specifically. The older Information Technology Act, 2000, and the “reasonable security practices” mandated under Section 43A, offer only general data-security obligations. Absent specific Bar Council guidance, Indian lawyers are left to extrapolate confidentiality obligations from general fiduciary principles articulated in cases such as R.D. Saxena v. Balram Prasad Sharma, which characterized the advocate-client relationship as one of the highest trust.
VI. Toward a Regulatory Framework
Drawing on the comparative experience above, several principles suggest themselves for an Indian regulatory response:
Data protection safeguards specific to privileged information. The DPDP Act’s general framework should be supplemented with sector-specific guidance, potentially issued jointly by the Bar Council and the Ministry of Electronics and Information Technology such as addressing the use of client data in AI tools.
A codified technology-competence duty. The Bar Council of India should consider amending its Rules to include a provision analogous to ABA Comment 8, making explicit that competence includes understanding the capabilities and limitations of AI tools used in practice.
Mandatory verification standards. Building on Mata v. Avianca and Park v. Kim, courts and bar associations could adopt a clear rule that any AI-assisted citation or factual assertion must be independently verified against primary sources before filing, with certification requirements akin to those already used for e-filing.
Disclosure norms for judicial use of AI. Following the controversy in Jaswinder Singh, clearer guidance is needed on whether, and how, AI-generated material may be referenced in judicial orders, to preserve the reasoned, precedent-based character of adjudication.
Bias auditing for algorithmic tools used in the justice system. Where AI tools inform bail, sentencing, or case-triage decisions, an independent audit mechanism, informed by the Loomis debate and India’s evolving jurisprudence on algorithmic accountability post-Puttaswamy, should be considered.
VII. Conclusion
The duty of competence has always been a dynamic, rather than static, obligation — expanding historically to encompass electronic discovery, cybersecurity, and now artificial intelligence. What Mata v. Avianca, Park v. Kim, and Jaswinder Singh v. State of Punjab collectively demonstrate is that AI does not excuse lawyers or judges from the foundational obligations of verification, candour, and independent judgment; if anything, it raises the stakes for their observance, since the errors AI produces are unusually persuasive in form. India’s regulatory framework, still largely silent on the specific challenges of generative AI, would benefit from the comparative lessons of U.S. ethics opinions and case law, while adapting them to the distinct institutional context of the Advocates Act and the Bar Council of India Rules. The technology will continue to evolve; the underlying professional obligation, that a lawyer must know, and be accountable for, what leaves their hands, will not.
THIS ARTICLE IS WRITTEN BY KAVYANJALI B S FROM ST. JOSEPH’S COLLEGE OF LAW
REFERENCES :
Statutes and Rules
Advocates Act, 1961 (India)
Bar Council of India Rules, Part VI, Chapter II
Digital Personal Data Protection Act, 2023 (India)
Information Technology Act, 2000, Sec. 43A (India)
ABA Model Rules of Professional Conduct, Rules 1.1, 1.6 (2023)
Federal Rules of Civil Procedure, Rule 11 (U.S.)
Cases
Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)
Park v. Kim, 91 F.4th 610 (2d Cir. 2024)
Jaswinder Singh v. State of Punjab, CRM-M No. 22496 of 2022 (P&H HC, Mar. 27, 2023)
State v. Loomis, 881 N.W.2d 749 (Wis. 2016)
K.S. Puttaswamy v. Union of India, (2017) 10 SCC 1
R.D. Saxena v. Balram Prasad Sharma, (2000) 7 SCC 264
Scholarly Sources
Cathy O’Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy (2016).
Richard Susskind, Tomorrow’s Lawyers: An Introduction to Your Future (2d ed. 2017).
Dana Remus & Frank Levy, Can Robots Be Lawyers? Computers, Lawyers, and the Practice of Law, 30 Geo. J. Legal Ethics 501 (2017).
Harry Surden, Machine Learning and Law, 89 Wash. L. Rev. 87 (2014).
4–6. ABA Standing Committee on Ethics and Professional Responsibility, Formal Opinion 512: Generative Artificial Intelligence Tools (2024).
9–10. NITI Aayog, National Strategy for Artificial Intelligence (2018); commentary in NUJS Law Review on judicial AI adoption.
Indian Journal of Law and Technology, commentary on Jaswinder Singh v. State of Punjab (2023).
Cathy O’Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy (2016).
