AI is coming to Malaysian legal practice. Who sets the rules?

The firms that decide how AI enters their practice will do better than the ones who let it happen to them.

Walk into almost any law firm in the Klang Valley today and you will find the same quiet reality. Associates are already using generative AI. Not because the firm rolled it out, and not because a partner approved a tool. They are using it because the drafting never stops, the chronology still has to be built by morning, and the tools are one browser tab away.

That is the real starting point for any honest conversation about AI in Malaysian legal practice. It is not arriving. It has arrived. The only open question is whether the firm shapes how it is used, or discovers after the fact how it was used.

The profession has already moved

This is no longer a debate about whether the technology matters. In March 2025, LexisNexis Southeast Asia published its survey of more than 400 lawyers and legal professionals in Malaysia and Singapore. Seventy per cent of respondents believed they would fall behind if they did not start using generative AI tools. Sixty-six per cent said they were already using them in their work, with lawyers in private practice showing higher acceptance than those in-house. Fifty-six per cent considered the impact on their firm or organisation to be transformative or significant.

The more telling figure is the one that sits beside those numbers. Only 48 per cent expressed confidence in their understanding of generative AI tools and how to use them. In other words: a majority are already using the tools, a majority fear being left behind, and fewer than half feel they actually know what they are doing.

Pressure is not coming only from inside the firm. Respondents also pointed to the organisations they serve (36 per cent), their peers (30 per cent), and their clients (16 per cent). Managing partners who are still treating this as a future agenda item are already behind the behaviour in their own offices.

The Bar did not ban it. It reminded us who is responsible.

The Malaysian Bar has not prohibited generative AI. It has done something more important, and easier to miss. It has reminded members that existing professional obligations still apply.

Circular No 342/2023, issued on 24 November 2023, set out the risks and precautions in using generative AI in the legal profession, with particular reference to ChatGPT. The Bar Council made clear that it does not endorse or regulate any particular GenAI tool. What it did say, plainly, is that legal practitioners remain bound by their duties to the profession, to clients, to the courts, and to other stakeholders. Efficiency is not a defence to a breach of those duties.

Circular No 242/2025, issued on 3 July 2025, supplemented that guidance. It noted that GenAI has moved from novelty to an important professional tool, now appearing in legal research platforms, contract review systems, client engagement tools, and internal knowledge management. It also set out the risks that travel with that shift: confidentiality, accuracy, accountability, data security, and the unauthorised practice of law. It warned, specifically, about hallucinated content, breaches of client confidentiality, and overreliance on unverified AI-generated legal analysis.

That is not a ban. It is also not a blank cheque. The Bar's position is that members may use these tools, but they remain responsible for the output. If a chronology is wrong, if a citation does not exist, if confidential facts have been typed into a consumer chatbot, the problem is not "the AI". The problem is the lawyer, and the firm that had no rules.

Courts in other jurisdictions have already sanctioned lawyers for filing AI-fabricated authorities. Those cases travel quickly through the profession for a reason. They are not about technology. They are about a failure to check.

Across the causeway, the direction of travel is even clearer

On 6 March 2026, Singapore's Ministry of Law published its Guide for Using Generative AI in the Legal Sector. Malaysian firms should read it, even though it is not our regulator. It is the nearest full statement of what a mature regional framework looks like.

The Guide rests on three principles. First, professional ethics: legal professionals remain ultimately responsible for all work products, and must exercise the requisite knowledge, skill, and professional judgment. Second, confidentiality: users must safeguard client data and remain mindful of how different models process and store information. Third, transparency: lawyers should consider informing clients of GenAI use, particularly where it is substantial, affects cost, or involves data-handling that may not meet the client's requirements.

None of this is exotic. It is the ordinary law of lawyering, restated for a new tool. The lawyer still signs the advice. The firm still holds the client relationship. The court still expects the authorities to exist.

Malaysia's Chief Justice, Datuk Seri Wan Ahmad Farid Wan Salleh, made the same point in April 2026: AI is a tool, not a replacement for human judgment. That is the right frame. The question for a boutique firm is not whether judgment still matters. It is whether the firm will put AI in a place where judgment can still be exercised, or leave it in a browser tab where it cannot.

Boutiques have an advantage the large firms do not

Large firms move slowly on technology because everything must clear committees, networks, and legacy systems. A 30-to-50 lawyer firm in Kuala Lumpur can decide something on a Tuesday and be running it by the following month.

That agility is a genuine competitive edge. Used deliberately, it lets a boutique set rules, pick a practice group, and prove a workflow before the market standard hardens around someone else's terms. Used carelessly, the same speed lets ungoverned tools spread through the firm before anyone has thought about confidentiality or accuracy.

This is the part that tends to get lost in the noise. The risk for a boutique is not that it will be "too small for AI". The risk is that it will be small enough for AI to enter without anyone noticing, and then large enough for the consequences to matter. A consumer tool used on a live matter is not a junior's private shortcut. It is a firm problem: client-confidential information potentially leaving the firm's control, and an accuracy problem sitting on top of a confidentiality problem.

Banning the tools does not solve this. A ban feels responsible. In practice it drives the behaviour underground, which is how you lose both visibility and the goodwill of the juniors who want to work with modern tools. You cannot prohibit a technology that fits in a browser tab. You can only decide whether it is governed.

Blanket permission fails in the opposite direction. "Use AI, just be sensible" sounds reasonable and means nothing. It leaves each person to guess what is confidential, what needs checking, and what may be typed into which tool. That is how an unchecked draft ends up in front of a client, or a court.

What "setting the rules" actually looks like

There is a version of adoption that is neither reckless nor paralysed. It does not require an innovation committee, a six-month RFP, or a firm-wide rollout on day one. It looks like the way a well-run firm has always tried something new.

Pick one practice group. Disputes and corporate are the usual starting points, because that is where the document work sits. Pick two or three workflows that everyone already dislikes doing: chronologies, due-diligence summaries, first drafts from the firm's own precedents, retrieval from past matters. These are high-volume, low-judgment tasks. A lawyer is going to review the output anyway. That is where AI assists without pretending to replace the advice.

Then put four rules around them. One approved tool, chosen because it keeps data inside the firm's own environment — e.g. Microsoft 365 or Google Workspace — and does not train on client files. Named use cases, and named exclusions. A review rule: nothing AI-assisted reaches a client or a court without a lawyer's sign-off. A record: matter isolation and an audit trail, so the firm can answer "what was used, and by whom" if the question is ever asked.

Run that for sixty days. See what actually changes. If it is not working, stop it. If it is working, expand it. That is not a leap of faith. It is a contained pilot.

The firms that will look strongest in three years are not the ones that adopted AI first, or the ones that banned it longest. They are the ones who adopted it on their own terms, before the choice was made for them by their associates, their clients, or a mistake that could no longer be walked back.

What is holding your firm back right now? The risk, the cost, or simply no time to look at it properly?

Auxilator helps boutique law firms in Malaysia adopt AI safely — inside the firm's existing tools, grounded in the firm's own precedents, with lawyers reviewing every output and no training on client data.

Notes

LexisNexis Southeast Asia, Generative AI and the Legal Profession 2025 Survey Report, Malaysia and Singapore (press release, 25 March 2025). Survey of over 400 lawyers and legal professionals. Figures used: 70% fear falling behind; 66% already using GenAI (higher acceptance in law firms than in-house); 56% transformative or significant impact; 48% confident in understanding the tools; pressure from organisations served (36%), peers (30%), clients (16%). https://www.lexisnexis.com/blogs/my/b/press-room/posts/generative_2d00_ai_2d00_survey_2d00_report_2d00_2025
Malaysian Bar, Circular No 342/2023 (24 November 2023), The Risks and Precautions in Using Generative Artificial Intelligence in the Legal Profession, Specifically ChatGPT. https://www.malaysianbar.org.my/cms/upload_files/document/Circular%20No%20342-2023.pdf
Malaysian Bar, Circular No 242/2025 (3 July 2025), Updates on the Use of Generative Artificial Intelligence in Legal Practice. Supplemental to Circular No 342/2023. https://www.malaysianbar.org.my/cms/upload_files/document/Circular%20No%20242-2025.pdf
Mata v. Avianca, Inc., No. 22-cv-1461 (PKC) (S.D.N.Y. 22 June 2023) (sanctions for filing ChatGPT-fabricated case authorities). Circular No 242/2025 also flags global examples of lawyers relying on unvetted GenAI output in litigation. https://www.courtlistener.com/opinion/7275776/mata-v-avianca-inc/
Singapore Ministry of Law, Guide for Using Generative AI in the Legal Sector (published 6 March 2026). https://www.mlaw.gov.sg/files/Guide_for_using_Generative_AI_in_the_Legal_Sector__Published_on_6_Mar_2026_.pdf — Launch statement (three principles): https://www.mlaw.gov.sg/launch-of-guide-for-using-generative-artificial-intelligence-in-the-legal-sector/
Ahmad Mukhsein Mukhtar, "Chief Justice: AI is a tool, not a replacement for human judgment", New Straits Times (2 April 2026). https://www.nst.com.my/news/nation/2026/04/1408709/chief-justice-ai-tool-not-replacement-human-judgment — Same speech reported in "CJ: Lawyers should treat tech as an aiding tool", The Star (3 April 2026). https://www.thestar.com.my/news/nation/2026/04/03/cj-lawyers-should-treat-tech-as-an-aiding-tool