Generative AI can help a law firm search material, draft a first version or organise a large case file. It can also produce confident mistakes or expose sensitive information if used carelessly. In March 2026, Singapore’s Ministry of Law launched a Guide for Using Generative Artificial Intelligence in the Legal Sector. The guide is directed at responsible professional use, but clients have a stake in the answer to a simple question: how is my matter being handled when AI is involved?
The Ministry identifies three principles: professional ethics, confidentiality and transparency. None means that a lawyer can outsource judgment to a chatbot. Nor does the guide say every minor use of AI requires a dramatic disclosure. It calls for careful professional decisions, especially where AI plays a substantial part in work, affects cost or raises data-handling concerns. Here is a client-focused way to understand the announcement.
What MinLaw released in March 2026
The Ministry of Law announcement says the guide was launched on 6 March 2026 after a public consultation in September 2025. More than 20 local and international stakeholders provided feedback, including law practices, in-house teams, legaltech providers and academics. The document contains practical guidance and case studies about research, drafting quality and knowledge management in practices of different sizes. It is guidance on responsible adoption, not a promise that an AI output is correct.
Clients should separate the technology from the professional service purchased. A lawyer may use a tool to locate issues quickly but remains responsible for checking law, facts and the suitability of advice. A polished document is not proof that authorities were verified. The guide’s value lies in making responsibilities explicit while firms experiment with new workflows. Our article on lawyer-client confidentiality explains the broader duty that still applies when technology changes.
Principle one: professional ethics and human judgment
MinLaw says legal professionals remain ultimately responsible for all work products and must apply the necessary knowledge, skill and professional judgment. For a client, that means there should be an accountable person who can explain the reasoning and correct an error. Ask who reviewed a draft, which primary authority supports a legal conclusion and whether factual assertions were checked against your documents. An answer consisting only of “the software generated it” is not a satisfactory professional explanation.
This matters especially for an urgent injunction, a plea, a filing deadline or a high-value transaction. A fabricated case citation or a missed qualification can cause practical harm regardless of how efficiently the first draft was made. A lawyer should assess the tool’s output within the law and facts of the matter. Clients can help by identifying inaccuracies in their own chronology and by providing complete source documents rather than a selective summary.
Principle two: protect confidential client data
The guide says users must safeguard client data and confidentiality and consider how different AI models process and store information. A firm should understand whether prompts or uploaded files leave its controlled environment, who can access them, how long they are retained and whether they can be used to improve a model. The answer may vary across products and settings. A client should not assume that every tool bearing the same brand has identical contractual and privacy controls.
Ask what categories of material may be entered into AI, whether names and sensitive details are removed where possible, and whether the firm has approved the tool. For a matter involving children, medical records, trade secrets or criminal allegations, the stakes are obvious. Even seemingly ordinary correspondence can reveal a legal strategy. Our guide to preserving digital evidence for a lawyer is useful when supplying original records securely; do not destroy originals after uploading copies to any platform.
Principle three: meaningful transparency
MinLaw says lawyers should consider informing clients about AI use, particularly where the tool substantially produces legal work, affects the cost of services or involves data practices that may not align with client requirements. This is a contextual principle, not a universal claim that each spelling suggestion needs a separate consent form. A clear engagement discussion could identify approved uses, the review process, restrictions on confidential material and how AI-assisted work will be billed.
Clients can ask these questions without assuming that AI use is automatically poor practice. A well-supervised tool may save time; a poorly supervised one may create more review work or risk. If a firm charges by the hour, ask whether the quoted fee reflects the actual work performed and what human checking is included. Our guide to the legal engagement letter helps put scope and fees in writing before the case moves forward.
Can a client say no to particular AI uses?
A client can communicate confidentiality or procurement requirements at the start and ask the firm whether it can meet them. Whether a specific restriction is accepted depends on the engagement, the firm’s systems and the work required. Put any agreed limitation in writing. For instance, distinguish a ban on uploading identifiable documents to an external model from routine internal search or spelling assistance. Vague instructions such as “no AI anywhere” can be hard to administer without defining what counts.
If the matter belongs to a business, coordinate legal instructions with its information security policy. Identify controlled data, cross-border transfer concerns and authorised tools. Ask how a breach or mistaken disclosure would be escalated. The PDPC’s AI personal data guidance provides broader context for organisations, although it addresses a different AI use case and does not replace the legal sector guide.
What to ask at your next lawyer meeting
Ask whether AI will be used to research, summarise evidence, draft advice or prepare court documents. Who approves its use and reviews the result? Will identifiable documents be uploaded to a third party? Will your information be retained or used for training? How will fees be calculated? How can you request correction if you spot an error? These questions are best asked before sensitive files are shared, when the firm can explain its actual workflow and agree a sensible boundary.
Keep the discussion proportional. A short routine matter may need only a brief statement of practice. A complex litigation or transaction may justify a detailed protocol. If your original lawyer leaves or a new team takes over, confirm that any special AI or confidentiality instruction travels with the file. The official MinLaw release links to the full guide for those who want the precise professional framework.
If you suspect a generated mistake
Point to the exact sentence, citation or factual statement and ask the supervising lawyer to verify it against an authoritative source. Keep the version you received and the underlying document that shows the discrepancy. Do not assume that a strange phrase proves AI was used; humans make errors too. The important response is correction, an explanation of any effect on advice already given, and prompt action if a filing or deadline is involved. If you remain concerned, request a second opinion with a clear account of what has been checked and what is still uncertain.
The guide’s practical message
The March 2026 development neither bans AI in legal practice nor turns a generated answer into legal advice. It asks professionals to maintain competence, protect confidential information and consider transparency when use is consequential. For clients, the useful response is to ask about review, data handling and cost in plain language. A firm should be able to describe how a qualified person remains accountable for the final work.
Technology will continue to change. The questions above remain durable because they focus on who checks the advice, where the information goes and what the client pays for. If a firm cannot answer those questions for a sensitive matter, pause before sending the next batch of documents and clarify the engagement.
Frequently Asked Questions
Does the 2026 guide prohibit lawyers from using generative AI?
No. It promotes safe and responsible use under professional ethics, confidentiality and transparency principles.
Must a lawyer disclose every use of AI to a client?
The announcement says lawyers should consider informing clients especially when AI use is substantial, affects cost or raises data-handling concerns. It does not state a blanket rule for every minor use.
Who is responsible if an AI-assisted legal document contains an error?
MinLaw says legal professionals remain ultimately responsible for their work products and must exercise professional judgment.
Thank you for sharing this article...
About the Author: Randy Alta
