CLA News / From the Grassroots to the Frontier: How AI Is Democratizing the Legal Profession By Ooi Wei Qi

Introduction: The Invisible Barrier

Access to the legal profession has never depended solely on formal legal education. In litigation particularly, competence is developed through exposure: observing advocates, translating principles into strategy, learning to draft, and making decisions under pressure. Much of this is not found in textbooks. It is acquired through proximity to those who have already navigated the profession.

I am the first lawyer in my family. Growing up in Penang, there were no legal texts on the shelves, no relatives to explain what it meant to be called to the Bar. When I entered practice, I quickly realized the gap between me and many of my peers was not one of knowledge, but of access:  access to mentors who understood the unwritten rules of practice, and to the tacit knowledge that rarely appears in textbooks but defines professional competence. For lawyers from established legal families, such knowledge is ambient. For first-generation lawyers, it must be actively sought.

This is the profession’s hidden curriculum. Artificial intelligence (AI) is beginning to disrupt this architecture of access—not as a substitute for mentorship, but as a new form of infrastructure for legal learning.  It cannot replicate professional judgment or the trust embedded in human relationships. However, it can reduce the exclusivity of access to foundational legal knowledge, giving lawyers who lack traditional networks a new means of entering the professional conversation.

In Malaysia, this remains a potential rather than an automatic outcome. The Malaysian Bar has recognized the practical value of generative AI while emphasizing that lawyers remain responsible for verifying AI-generated material.[1] That caution is necessary, but incomplete: the deeper question is whether lawyers across different practice environments can realistically access and adopt AI at all. A sole practitioner has fewer institutional precedents and less structured training than a lawyer in a large firm. AI could narrow that disparity but only if reliable tools and training are themselves accessible. Otherwise, a technology capable of flattening professional hierarchies may simply reinforce them.

The Mentorship Gap: Law’s Hidden Curriculum

Across common law jurisdictions, legal training has long been grounded in apprenticeship: law school provides doctrinal foundations, but competence is built through incremental responsibility under supervision.[2]  That system assumes consistent access to mentorship. In reality, such access is unevenly distributed, and first-generation lawyers frequently navigate early practice without stable supervision.

In jurisdictions such as Malaysia and India, competence still depends heavily on informal apprenticeship: how a particular judge prefers arguments framed, how a registry processes filings, how procedural ambiguity is navigated.[3] These are forms of tacit knowledge acquired through proximity to experienced practitioners. AI cannot replicate that judgment, but it can provide immediate, context-specific explanations, drafting examples, and pathways to relevant authorities at the point of need, reducing the extent to which professional development depends on geography, firm size, or personal networks.

The AI Revolution: Access Without Authority

For first-generation lawyers, AI represents access rather than mere efficiency. I still remember preparing my first unfamiliar application. My instinct was not that I lacked ability, but that I lacked someone to ask whether I was approaching it correctly. AI did not replace a supervising lawyer. It provided a starting point, a way to organize the issues and identify the questions I still needed to answer myself.

Accessibility must not be confused with authority. AI output is a hypothesis to be tested, not a source of law, and the consequences of skipping that verification step are no longer theoretical. In the UK, the Divisional Court’s combined judgment in R (Ayinde) v London Borough of Haringey[4] and Al-Haroun v Qatar National Bank[5] is instructive on two distinct failure modes. In Ayinde,[6] a pupil barrister put a wholly invented case before the court, one of five fake authorities in her submissions. The Divisional Court’s decision not to pursue contempt turned partly on unresolved questions about whether she had been adequately supervised during pupillage. The AI failure was, on the court’s own account, difficult to separate from a mentorship failure. In Al-Haroun,[7] of the forty-five authorities put before the court, eighteen did not exist, and most of the remainder were misquoted or irrelevant. The hallucinated citations originated not from the lawyer’s own research but from the lay client, and the solicitor’s failure lay in relying on client-supplied material without independent verification.

Singapore’s courts have gone further in pinpointing who bears responsibility when AI errs, and in refusing to let the inquiry turn on whether AI use can actually be proven. In Tajudin bin Gulam Rasul and another v Suriaya bte Haja Mohideen,[8] the advocate was ordered to pay S$800 under a personal costs order for citing a fictitious AI-generated authority. When he attempted to attribute the error to a junior colleague who had “run it through an AI app,” the court gave that explanation no weight, holding that the senior lawyer on the file bore a non-delegable duty to supervise, regardless of who had actually generated the fabricated citation. A personal costs order of S$900 followed in Goh Chin Cheng v Choco Up SG Pte Ltd.[9] Crucially, this was not a case involving an AI-generated fictitious authority. Counsel denied using AI; instead, the issue was the failure to verify and the serious misquotation of genuine authorities. Counsel had attributed a fictitious block quotation and directly contradictory holdings to two real cases. The court held that it was immaterial whether these errors resulted from AI or human carelessness. The sanction arose from counsel’s failure to discharge the non-delegable duty to verify the accuracy of the materials placed before the court.

Weeks later, a High Court judge issued personal costs orders in Tan Hai Peng Micheal and another (as the executors of the estate of Tan Thuan Teck, deceased) v Tan Cheong Joo and another and other matters,[10] ordering each of the two lawyers on the file to personally pay S$5,000. These were not merely compensatory orders; they reflected the court’s strong concern with professional responsibility and deterrence. This was the first published decision of a judge of the General Division of the High Court dealing substantively with the making of personal costs orders against counsel for fictitious authorities likely generated by an AI tool. Malaysia, notably, has yet to produce a reported court decision on this issue. For now, lawyers there are guided mainly by the Bar Council’s guidance rather than by court decisions, leaving a gap in the law that should be acknowledged rather than overlooked.

These cases raise an important question that the legal profession has been slow to address: is a lawyer now expected to understand technology sufficiently to know when information produced by AI requires verification, just as lawyers are already expected to protect confidentiality and exercise independent judgment? The Singapore cases suggest the answer is trending toward yes, and that this duty cannot be delegated downward to a junior or outward to the tool itself. This is also where the distinction between AI assisting a lawyer and AI advising a member of the public directly becomes important: the professional-development story this article tells concerns the former, where a supervising lawyer remains accountable for the output. The latter, where AI substitutes for legal advice rather than augmenting it, raises a different set of access-to-justice questions this article does not attempt to resolve.

Across common law jurisdictions, professional bodies are converging on the same underlying position without eliminating the divide between well-resourced and under-resourced practices. The Singapore Academy of Law[11] and the Law Society of Singapore[12] have issued structured guidance. Courts and professional bodies in the United Kingdom have issued comparable warnings.[13] The difference between jurisdictions is not whether AI is permitted, but whether the profession and the state are helping lawyers acquire the capacity to use it responsibly.

This transition carries a deeper risk than individual sanction. AI is displacing the mechanical work of junior practice, including research, document review and first drafts, while promising to redirect juniors toward verification and strategic judgment.[14] Nevertheless, judgment is not a function into which juniors simply graduate. It accumulates through the friction of performing the very routine tasks AI now automates.[15] A junior who has never wrestled with a flawed chronology may lack the tacit knowledge to recognize the fatal error in an AI-generated one. The danger is not only that entry-level roles shrink, but that the apprenticeship ladder itself hollows out.[16] AI may thus democratize access to legal knowledge while narrowing access to the professional opportunities through which that knowledge becomes judgment.

Doctrine, however, is only one dimension of competence. Court craft, including reading the bench, adjusting arguments in response to judicial intervention and managing the flow of a hearing, has always required a different kind of apprenticeship, one that AI cannot reproduce. This is where the parallel expansion of digital access to court proceedings matters: livestreaming by the Supreme Court of India[17] and the Supreme Court of the United Kingdom,[18] together with Malaysia’s growing digital court infrastructure,[19] lets a first-generation lawyer in a district town observe how experienced counsel argue without requiring physical access to a senior practitioner’s chambers. AI can then transcribe and structure those proceedings, turning observation into material for reflection. Neither replaces mentorship or courtroom experience, but together they loosen professional learning’s historical dependence on inherited proximity.

The New Frontier: Collaboration Over Pedigree

If AI erodes the informational advantage that elite institutional networks have historically conferred, it forces a redefinition of what makes a lawyer valuable. When any lawyer can generate a competent first draft or identify potentially relevant precedent in seconds, differentiation can no longer rest on who has the precedent. It must rest on what a lawyer does with it: verifying the authority, discerning its limits, and exercising judgment when the law is unsettled.

For lawyers seeking to develop professionally, disadvantage has rarely been determined by background alone; but also, by proximity to major courts, well-resourced firms, and professional communities where reputations are built through visibility. A lawyer practicing in Penang may possess the same ability as a practitioner in Kuala Lumpur, yet face fewer opportunities for daily exposure to complex appellate argument. AI, combined with livestreamed proceedings, begins to loosen that constraint: a lawyer who cannot attend a Bar Council seminar in Kuala Lumpur can use AI-assisted research to engage with unfamiliar doctrine; one who has never observed appellate advocacy in person can watch Federal Court proceedings online. Professional development becomes less dependent on inherited networks and more on analytical discipline and initiative. It reshapes inequality rather than eliminating it. Hierarchy is not dismantled but reconfigured around who can collaborate effectively with artificial intelligence.

The Commonwealth Imperative: Responsibility and Regulation

Across the Commonwealth, AI-driven platforms are increasingly used to provide legal information and support legal aid systems where access to lawyers is limited.[20] That is a genuinely different question from the one this article addresses: not whether AI can widen access to legal services, but whether it can widen access to legal careers. The two are connected but not identical, and conflating them risks understating how much regulatory work remains to be done on each.

AI itself is not equally accessible. Subscription costs, digital literacy, and organizational data-governance requirements may create new forms of inequality even as older barriers fall. For Malaysia specifically, the absence of a reported case leaves open several questions that Singapore and the UK have already begun to answer: whether supervising partners bear non-delegable responsibility for a junior’s unverified AI use, whether existing professional-conduct rules need amendment rather than mere guidance, and whether the courts here will treat a first Malaysian incident with the leniency shown in early UK and Singapore decisions or the escalating severity seen as the law has developed elsewhere. The profession would be better served addressing these questions before a case forces the issue.

Conclusion: Towards an Inclusive Legal Future

I entered the profession without inherited networks or informal access to mentorship. AI does not eliminate that imbalance, but it alters the structure of early-stage professional learning: providing immediate explanations, reducing dependence on proximity to institutional knowledge, and substituting for some of mentorship’s more accessible functions without displacing mentorship itself.

For young Commonwealth lawyers, this carries both opportunity and responsibility. The opportunity lies in using AI to accelerate learning and broaden participation. The responsibility lies in using it critically, with full awareness of the duty to verify its output—a duty that courts across the Commonwealth are now actively defining, but that Malaysia has not yet had to test.Success for the next generation may depend less on who first opens the door, and more on what they do, and how carefully they check their work, once they walk through it.

Ooi Wei Qi

Advocate & Solicitor, Jeeva Partnership, Malaysia

[1] ‘Updates on the Use of Generative Artificial Intelligence in Legal Practice’ (Badan Peguam Malaysia Malaysian Bar, 3 July 2025) <https://www.malaysianbar.org.my/cms/upload_files/document/Circular%20No%20242-2025.pdf> accessed 15 August 2026.

[2] ‘Tim Perry Discusses the Challenges Facing Lawyers on the Journey from Graduate to Partnership [Q&A]’ (Thomson Reuters, 1 September 2017) <https://insight.thomsonreuters.com.au/legal/resources/resource/tim-perry-practical-law-australia-graduate-to-partnership-qanda> accessed 26 June 2026.

[3] Shraddha Gopal Chughra, ‘Building Strong Foundations – Reforming Early Careers for Junior Lawyers in India’ (SSRN, 30 December 2025) <https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5879803> accessed 15 August 2026.

[4] R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin).

[5] Al-Haroun v Qatar National Bank [2025] EWHC 1588 (Comm).

[6] Ayinde (n 4).

[7] Al-Haroun (n 5).

[8] Tajudin bin Gulam Rasul and another v Suriaya bte Haja Mohideen [2025] SGHCR 33.

[9] Goh Chin Cheng v Choco Up SG Pte Ltd [2026] SGHCR 13.

[10] Tan Hai Peng Micheal and another (as the executors of the estate of Tan Thuan Teck, deceased) v Tan Cheong Joo and another and other matters [2026] SGHC 49

[11] ‘Prompt Engineering for Lawyers’ (Singapore Academy of Law, 2024) <https://sal.org.sg/wp-content/uploads/2025/02/SAL-Microsoft-PE-Guide.pdf> accessed 26 June 2026.

[12] ‘Law Society’s Advisory on the Use of Publicly Available AI Tools’ (The Law Society of Singapore, 2 April 2026) <https://www.lawsociety.org.sg/wp-content/uploads/2026/04/Law-Societys-Advisory-on-the-Use-of-Publicly-Available-AI-Tools-2-April-2026.pdf> accessed 26 June 2026.

[13] ‘Generative AI – the essentials’ (The Law Society, 1 October 2025) <https://www.lawsociety.org.uk/topics/ai-and-lawtech/generative-ai-the-essentials> accessed 15 August 2026.

[14] Angela Tufvesson, ‘Is AI eroding the legal apprenticeship?’ (LSJ Online, 13 July 2026) <https://lsj.com.au/articles/is-ai-eroding-the-legal-apprenticeship/> accessed 16 August 2026.

[15] Lachlan Robb, Rachel Hews, Felicity Deane, Michael Guihot, Jonah Farry and Amanda Kennedy, ‘It’s the End of the World as We Know It’: Generative Artificial Intelligence and the Changing Landscape of Legal Practice and Education’ (2025) 48 (4) UNSW Law Journal 1129.

[16] Anthea Roberts and David Wilkins, ‘The Training Crisis’ (Center on the Legal Profession Harvard Law School, 7 April 2026) <https://www.dragonflythinking.com/reports/ai-and-the-legal-profession/articles/?article=training-crisis> accessed 15 August 2026.

[17] ‘Live Streaming’ (Supreme Court of India, 14 August 2026) <https://www.sci.gov.in/live-streaming/> accessed 16 August 2026.

[18] ‘Latest judgments’ (The Supreme Court of the United Kingdom) <https://supremecourt.uk/news/latest-judgments> accessed 16 August 2026.

[19] Nordiana Barka & Nor Fariza Mohd Razli, ‘e-Kehakiman System Enhances Court Efficiency, Expands Public Access’ (Bernama, 21 December 2025) <https://www.bernama.com/en/news.php?id=2505009> accessed 16 August 2026.

[20] ‘AI and access to justice’ (The Commonwealth) <https://thecommonwealth.org/ai-and-access-to-justice> accessed 27 June 2026.