How Does AI Impact The Legal Profession And Its Ethical Standards

Author: LawSphereX

Category: Technology & Cyber Law

Published: 2026-08-23T20:14:24.118572+00:00

Updated: 2026-08-26T17:11:18.789+00:00

HOW DOES AI IMPACT THE LEGAL PROFESSION AND

ITS ETHICAL STANDARDS?

Pavithra Jayan, Manipal Law School, Manipal Academy of Higher Education

ABSTRACT

The integration of artificial intelligence into the legal field is one of the most significant revolutions the profession has undergone in a while. While AI and its rapid growth promise efficiency, it also raises multiple ethical and regulatory challenges which the current legal systems are unprepared to address and navigate. There are multiple ethical conflicts that generative AI creates – questions about competence, model hallucinations, unsupervised and unreviewed data, potential bias and more. There is a gap in the accountability and all the impacts it has on third parties (especially the litigants). This paper, based on judicial decisions, scholarships and jurisdictional professional conduct rules, argue that most legislations that exist, as of now, are structurally inadequate to protect the clients, ensure proper AI usage and uphold the duties of the legal professionals.

Introduction Artificial Intelligence (AI) and its usage in the legal profession have been debated for decades now, starting with a conference in 1956, soon moving to machine learning and predictive analysis forms of AI and, in the past few years came generative AI, the most controversial tool in the current times. Still in its developmental stages, the regulation is hugely minimal and hence, consequences still being discovered. From basic drafting, comparisons and analysis to document research and reviews – AI has been an easy solution to professionals all around. 1 Though the potential of this technology is vast and new, the implications also remain so – there are questions about competence, bias, confidentiality, and accountability that AI cannot answer.

The existing ethical standards were created and enforced when the legal profession was still mostly manual and when the services were delivered with the professional judgements of real individuals. The Bar Council - Rules on Professional Standards2, The American Bar Association (ABA) Model Rules of Professional Conduct3 and more, were all drafted with the intention of guiding advocates to behave professionally with the court, the clients and the society. The growth of AI is such a manner was not predictable and applying the same Codes hence, do not make sense. This essay attempts to address the issues with unchecked AI usage in law and its ethical conundrums, trying to argue that while AI does provide speed and ease to professionals from all branches, many of these fields need human interpretation and learned opinions, which technology cannot replicate yet.

A machine or a system may do the work faster than a human, but speed is not the reason law is considered to be a noble profession. The Current Usage of AI in Law AI is already becoming an indispensable tool in research and drafting, among everyone in the legal world from law students to law firm partners. It is already leading to new ethical conflicts between professors and students, with the life of the latter being simplified, while the former tries to impart the value and necessity of actual work and knowledge. While ChatGPT and 1Robert J. Couture, ‘The Impact of Artificial Intelligence on Law Firms’ Business Models’ (Center on the Legal Profession, Harvard Law School, 24 February 2025)

2 The Bar Council of India, Rules on Professional Standards

3 The American Bar Association (ABA), Model Rules of Professional Conduct (1983)

OpenAI are easily accessible and affordable, there has been a growth in legal specific language models that are trained to focus better4 – as an answer to the demands of those professionals who spend nights in the office trying to classify and review numerous documents. LawLens, CoCounsel and Legora are all examples of legal tech AI that help draft contracts, provide support with arguments and drafts and even review your documents and provide alternatives to problematic clauses. They also cite authorities, precedents and identify certain patterns in the ratios of judgements. These models also help students and researchers summarize cases and simplify complex legal terms, making it more understandable. Even the Supreme Court of India has its own AI model – SUPACE to help with the overwhelming number of cases.

5 It is designed specifically to process and structure information and hence does not interfere with the judicial reasoning. Legal tech may also be of use to non-lawyers who may generate letters and pleadings, indistinguishable from experts which could potentially be a problem, if not reviewed properly. This is where AI has its limitations. Competence of AI and Competence of the Lawyer The fundamental duty of a lawyer is to make sure they are competent in their job – in this instance, competence is measured by how well versed and experienced they are, in the particular type of case they are dealing with. This is also what the client trusts and expects from the lawyer. The ethical question here is when AI is used for the wrong reasons, not the total usage by itself.

This is because the other dimension of competence is being aware and to keep abreast of all new developments in the field, including the changes in legal tech (AI). 6 ABA Rule 1. 1: ‘competence’ is also the obligation of lawyers to know the benefits and risks of said developments, and later it was made to include ‘tech competence’ requiring the knowledge of data privacy and electronic communication, so as to improve the forms of confidentiality in the sector. 7 While requiring ‘legal tech’ expertise, it is important to understand the different issues that 4 Ryan McCarl, ‘The Limits of Law and AI’ (2022) 90 University of Cincinnati Law Review 928, 928–930.

5 Ministry of Law and Justice, ‘Use of AI in Supreme Court Case Management’ (Press Information Bureau Delhi, 20 March 2025) 6 Miles Indest and others, ‘Legal Profession, Professional Perspective - Tech Competence From the Courtroom to Cyberspace’ (McGuireWoods LLP, May 2023) 7American Bar Association, ‘Rule 1. 1 Competence’ Model Rules of Professional Conduct (https://americanbar. org/groups/professional_responsibility/publications/model_rules_of_professional_ conduct/rule_1_1_competence)

grow along with technology. The matters in which one can blindly rely on AI is limited. Lawyers need to first understand how the tools work, the mode of use, the reliable outputs and how to verify them – for all of this, human judgement is absolutely a non-negotiable. There is also an obligation on them to know when these tools can be inappropriate. AI is trained in certain matters, and many platforms are also trained to not reply negatively or with an ‘I don’t know’ – this means that the platform may generate citations that have a word similar to the given prompt, but has no connection whatsoever to the actual matter at hand. They can also generate non-existent cases sometimes. These non-existent cases are called ‘phantom precedents’ or ‘AI hallucinations’.

8 While sounding extremely relevant and plausible, they are incorrect answers or cases fabricated when an LLM (Large Language Model) fill in gaps where they are inadequately trained or due to misinterpretations and bias. 9 This is also because generative AI runs on the model of ‘probability’ and so, it triggers whichever is the easiest and most likely response, without checking for accuracy. When it is left unreviewed and used in briefs and other documents, it risks harm to the lawyer’s reputation, the case and even to the client’s life, thereby affecting everything it was supposed to help. In Mata v Avianca, INC (2023), the lawyers of the plaintiff used ChatGPT to draft the legal brief. The court couldn’t find the cases cited and it was discovered that most of them were non-existent and the document wasn’t reviewed before submitting.

10 There are also many cases which had similar issues – in a land dispute case in Andhra Pradesh, the lawyer cited multiple judgements, that were later found to be fabricated by ChatGPT. 11 Though the lawyer claimed to have made the mistake in ‘good faith’, the court held that the lawyer’s duties are fundamental principles in the justice system. There are also multiple cases in Bombay, where similar hallucinated citations have been added into the briefs. In these situations, the lawyers are making a systemic choice to prioritize speed over responsibility. The Principle of Confidentiality All the jurisdictions around the world follow the principle of client-lawyer confidentiality.

It 8 Athaluri SA and others, ‘Exploring the Boundaries of Reality: Investigating the Phenomenon of Artificial Intelligence Hallucination in Scientific Writing Through ChatGPT References’ (2023) 15(4) Cureus e37432.

9 Özer M, ‘Is Artificial Intelligence Hallucinating?’ (2024) 35(4) Turk Psikiyatri Dergisi 333–335.

10 Mata v Avianca Inc 678 F Supp 3d 443 (SDNY 2023)

11 Vineet Bhalla, ‘AI “Hallucination” Again in a Court Order, SC Talks of “Institutional Concern”’ The Indian Express (2026).

may be practiced in various manners, but the core remains the same. 12 Law students are taught the importance of keeping the secrets of their clients and how being a professional means to take them to their graves. It is an enforceable ground and a form of protection. The use of technology, especially AI, can be detrimental in this case as sensitive data gets transmitted from lawyers to clients and third parties daily – while not knowing whose platforms process, store and train these AI models and how much access they may have to the uploaded data. This is extremely plausible in the Indian context because the majority of AI platforms used are either based or funded in the United States, and the data uploaded is stored on foreign platforms that are not easily traceable. This is severely misunderstood by all parties involved and the risks to relying on unchecked platforms are immense.

Digital footprints can never be removed, which means that once the data is out, it may continue to exist even after being removed from websites, as there’s always a possibility that any random person may have saved copies or screenshots. Many of the documents uploaded may also be used to improve the AI model while training them for future performance – these are all questions that arise directly with regard to confidentiality principle or the fiduciary nature between clients and lawyers. 13 Sometimes, even lawyers, who are literally trained to go over and scrutinize every word in contracts, accept all ‘cookies’ and agree to the terms and conditions given, without a second thought. This too may drastically affect the usage of data provided to the model.

It is another structural consequence of unregulated usage of random AI models along with the high chance of losing attorney-client privilege. The same way any contract can indemnify a party against all issues that could arise, digital acceptance – the ‘I agree’ button, in these situations, can also prohibit one from bringing a case against the service or the creators. In many jurisdictions, this is seen as a binding contract in the eyes of the law. The implication is that the user has granted rights to the platform and its employees to use the data, store it and access the same if necessary. Risks and Accountability The risks related to improper usage of AI form one of the most serious ethical challenges. AI systems are trained on the data that is fed to it. This information may consist of patterns and

12 Nicole Engler, ‘Using AI In Legal Work: COPRAC’s Tips On Confidentiality And Competence’ (24 May

2024)

13 Aduragbemi Odubela, ‘AI, Client Secrets, and the Lawyer's Duty of Confidentiality’ (Leeds Beckett

University, 2025)

historical issues and the model will tend to form its opinions or answers based on the same. This causes bias in all the information generated later. When the technical team/person themselves are biased, there is a chance that models can be trained to respond to certain queries in the way they think. For example, in a US case, State v Loomis, the judges used COMPAS (AI) to predict the likelihood of guilt using historical and personal data. 14 In a country like US where racial bias in prisons run high, it is very likely that black defendants could be labelled guilty, just because of group statistics provided. Leading to uneducated guesses and systemic unequal patters in the answers, it may also cause huge risks to the lawyer who decides to trust the system and not review the same. Transparency in AI tools is very restricted.

Many of these models and their mechanisms are protected under intellectual property laws and no one knows the manner in which they reach their conclusions and what proof it holds to back the claims. 15 This is called the ‘black box problem’. It interferes with fair procedure and the foundation of the adversarial system itself. With even judges using predictive tools, all possible harm falls on the clients. The disclaimer clauses and the limitation caps that these platforms use ends up holding the user liable for not fact checking. This means that if the platform, due to its addition of the technical clause, wins the case, the individual may even have to cover their legal costs. Many AI models also include clauses that allow only arbitrations and prevent class action suits, thereby reducing their risks as many users won’t even be able to afford the same.

Courts and committees have often debated about the challenges in using AI and while acknowledging the ease it provides, they also highlight the limitations. The New York City Bar Formal Opinions – Professional Ethics Committee - have even made guidelines and listed ethical obligations of their lawyers and firms using generative AI in practice. 16 This does not mean that all the responsibilities of the AI company are waived off, but it helps lawyers work in a manner that does not harm the client while still keeping up with the legal tech. Regulation and Justice Like any new technology, AI has its pros and cons. And like when any new technology

14 ‘State v Loomis’ (2017) 130 Harvard Law Review 1530

15 Dr. Juris Vasiliki Papadouli, ‘Transparency in Artificial Intelligence: A Legal Perspective’ (2022) 4(1) Journal of Ethics and Legal Technologies 27. 16 New York City Bar Association, Formal Opinion 2024-5: Generative AI in the Practice of Law (7 August 2024).

develops, the debate about its benefits and consequences has also developed in the legal field. One of the most significant achievements of AI is making basic laws accessible and understandable to those who struggle to afford it. 17 AI legal assistance can provide anyone with basic level of guidance in the least, especially when they would otherwise be absolutely unaided. 18 In many jurisdictions, apps and AI tools (DoNotPay, Lawful. ai) already exist legally which help citizens challenge parking/speeding fines, appeal executive decisions and other basic rights-based sensitisation without a lawyer. This, if improved and built on, may be a promising push towards access to legal service as phones are currently a necessity and can be found in almost every person’s hand.

Even in India, ‘NyayGuru’ is an AI powered legal assistant who answers basic questions and connects them to specific legislations in Hindi or English. The Supreme Court of India has released a White Paper - “Artificial Intelligence and the Judiciary”19, to ensure that the usage of AI, which is inevitable, is fair and transparent and to also emphasize that it should be for enhancing reasoning and not replacing. But it has no formal binding value, which means that advocates are free of regulation. With millions of enrolled advocates and similar number of cases, a judiciary managing cases across various economic and geographic scales, India has a lot to gain from AI. All it needs is proper regulation and training.

In the US, the American Bar Association has made multiple remarks and opinions about AI, especially in the context of confidentiality and accountability but there is no specific binding guideline as of now. To assist judicial officers in the proper usage of AI, “Artificial Intelligence (AI) – Judicial Guidance (October 2025)” was published, replacing the April version, applicable in England and Wales. 20 The lack of governance here is concerning. There is always a lag whenever new technologies come into use and historically too, the legal system itself has been slow to adapt to these changes. The system, similarly has been delaying the control of AI by treating the problem as more technical than ethical, but not regulating is also passive permission.

21 But with the pace 17 Victorian Law Reform Commission, Artificial Intelligence in Victoria’s Courts and Tribunals: Consultation Paper, Benefits and Risks of AI. 18 The Mississippi Bar, AI Tools for Lawyers – A Practical Guide (Law Practice Management and Technology Committee, 2025). 19 Dr. Anurag Bhaskar and others, White Paper on Artificial Intelligence and the Judiciary (Centre for Research and Planning, Supreme Court of India).

20 Courts and Tribunals Judiciary, Artificial Intelligence (AI) – Judicial Guidance (October 2025).

21 Michael Kirby, ‘The Fundamental Problem of Regulating Technology’ (2009) 5 The Indian Journal of Law and Technology 14, 14–16

AI is developing and being used, lack of some sort of immediate regulation will grant lawyers careless use of powerful and potentially harmful tools. This will be even more difficult to regulate or remove once AI content is deeply embedded into the workflow itself. Many jurisdictions have addressed the fact that fake cases and wrong statutes have been cited multiple times due to the dependency of lawyers on AI. This could lead to a broken foundation for new advocates and also, lack of confidence among the citizens in the system. By penalizing lawyers alone, the problem may not get fixed.

The better way to adopt would be to explicitly acknowledge the use, make binding rules with respect to the use being in accordance with the already established principles of confidentiality, transparency and competency and by creating accountability mechanisms in case of any kind of harm caused due to the AI usage. Conclusion Artificial Intelligence is not a new development in the legal profession, nor is it a new challenge for the field to address. The skills of a professional advocate, the most fundamental duties, competence, protection of confidential information, commitment to justice and a passion to win were the most human characteristics that defined them. AI disrupts the illusion of a natural person standing at the centre of every legal matter with individual judgement and reasoning, and most importantly humane emotions.

These are basic qualities that lawyers learn and cultivate alongside legal knowledge that AI cannot replicate. In most firms, junior lawyers are constantly supervised and their work is reviewed but AI tools do not go through the same. The justification is that the point of AI is to increase efficiency so as to waste less time on review – but this is a flawed move, when the tool itself gives a warning of possible errors. There should also be a mechanism to distribute responsibility, when AI causes harm, that does not solely focus on the user (advocate). The platform’s waiver of all claims due to a disclosure is unjustified. If the creators have taken upon themselves to make a tool specifically for legal purposes and makes it commercially accessible to the public, well aware of all the possible faults, they should also be called upon to bear the risks. Hence a need for shared liability.

This is already being discussed and applied to malpractice suits in case of AI assisted clinical care. 22 This framework would require the platforms and providers to disclose basic information

22 Daniel Cham, ‘Navigating Shared Liability in AI-Assisted Clinical Care: A Comprehensive Guide for

Healthcare Professionals’ (September 2025).

including sources, limitations and the reasons for the assumptions or conclusions. With reference to each ethical principle, AI poses a problem – technical competency also means that the lawyer knows the limitation of the tools they use, duty of confidentiality exists in the digital sphere too and a promise of fairness means that lawyers and judges are able to identify bias and incorrect statements that the system could easily make, before coming to conclusions. With the adoption of these steps, AI usage can be self-regulated and much more efficient. Chapter II of the EU AI Act prohibits certain systems that could potentially cause more harm than aid. 23 Though it is a general framework on good AI guidance, many of its rules are applicable to the legal field and can be read with the ethical duties of a legal professional.

Similar binding provisions can be added to legislations in almost every jurisdiction to make a significant move towards regulating the proper usage of AI in legal and other fields. This is not just a desirable move, but necessary for professional integrity and access to justice.

23 European Parliament, Artificial Intelligence Act P9_TA(2024)0138 (2019–2024)