Singapore: AI use and professional conduct for lawyers
By the ChamberOne team 4 minute read
Singapore has been among the more active jurisdictions in publishing guidance on generative AI in legal work. The Ministry of Law launched a Guide for Using Generative AI in the Legal Sector on 6 March 2026, following a public consultation in September 2025. The guide sits alongside the professional conduct framework that already governs lawyers. This article summarises the Ministry's guide in outline, based on the official launch page, and suggests how a lawyer or small firm might work with it. It is general information, not advice. Read the full guide and the current conduct rules yourself.
Key points
- The Ministry of Law's guide was launched on 6 March 2026 and is framed as guidance; read the full text and the conduct rules yourself.
- Lawyers remain ultimately responsible for AI-assisted work products.
- Safeguard client confidentiality by checking how each tool processes and stores data.
- Consider telling clients about substantial AI use, especially where it affects cost or data handling.
- Turn the principles into a short firm policy with an owner, training and a review date.
What the guide is
According to the Ministry of Law, the guide promotes safe and responsible use of generative AI across the legal services sector. Besides setting out principles, it gives practical implementation guidance and case studies from law practices of different sizes and in-house legal teams.
The launch page describes it as guidance. It does not state that following it is mandatory, and this article does not say whether any part of it carries disciplinary weight. For how your conduct obligations interact with it, read the guide itself and the guidance from the Law Society and the relevant regulator.
Principle: responsibility stays with the lawyer
The Ministry's page says that legal professionals remain ultimately responsible for all work products, and that they must still apply the required knowledge, skill and professional judgement. That is the same message that appears in most guidance elsewhere, and it has a simple practical consequence: a lawyer cannot point to a tool as the reason an error reached a client or a court.
In daily practice, that means reviewing every AI-assisted output before relying on it, verifying every authority, and keeping the lawyer who signs the work in control of it.
Principle: confidentiality
The guide asks users to safeguard client data and confidentiality, and to consider the risks of different generative AI models, including how data is processed and stored. That points to questions any firm can ask before using a tool on client material.
- Is client input stored by the provider, and for how long?
- Is it used to train or improve models?
- Where is it processed, and which staff of the provider can see it?
- Can we limit use to approved tools and block personal accounts?
- What would we tell a client if they asked where their documents had gone?
Principle: transparency with clients
The Ministry's page says that legal professionals should consider informing clients about their use of generative AI, particularly where that use is substantial, affects the cost of legal services, or involves data handling that may not match client requirements. Note the wording: it asks the lawyer to consider, and it points to circumstances where telling the client matters most.
A practical approach is to put a short, plain statement about AI use in the engagement letter, and to raise it in conversation where a client has particular sensitivities, for example a client in a regulated industry with its own rules on outsourcing.
Turning the principles into firm practice
Suppose a five-lawyer Singapore firm wants to act on these themes. It might start with a one-page policy covering approved tools, what may be entered, who reviews output and when clients are told. It would assign a partner as owner, train staff in a short session, and review the policy after six months or when the guidance changes.
It would also keep a simple record for each matter where AI played a material part: which tool, which task, who checked it. That record helps if a client later asks how work was produced, and it supports supervision.
- 1Write a one-page policy
- 2Assign a partner as owner
- 3Train staff in a short session
- 4Record which tool, task and checker on matters where AI played a material part
- 5Review after six months or when guidance changes
- Then back to step 1.
Court filings and court-specific guidance
The Ministry's guide is aimed at the legal sector as a whole. Submissions and evidence in court raise their own questions, and the courts may publish separate guidance or practice directions for court users. If you appear in court, look for the current court guidance directly and check whether it applies to the forum and the kind of document you are filing.
The basic discipline is unchanged whatever the forum. Every authority cited must be found and read by a person, every quotation checked against the source, and the lawyer on record answers for the content. If a tool suggested a case, treat it as a lead to be verified, not as a source.
Keep a short note on the file showing how an AI-assisted submission was checked. If the court ever asks, the note answers the question in a minute.
- Have a person find and read every authority
- Check every quotation against the source
- Keep a note on the file showing how it was checked
- Look for the current court guidance directly
- Treat a case a tool suggested as a source rather than a lead
- Assume the Ministry's guide answers court-specific questions
In-house lawyers and smaller practices
The guide is aimed at the sector broadly, including in-house teams. An in-house lawyer has a further layer of internal policy, such as the employer's rules on data and approved tools. The two must be read together, and if they conflict the more cautious reading is sensible until you have clarified it.
Smaller practices sometimes assume that guidance is written for large firms. The case studies the Ministry mentions are drawn from practices of different sizes, which is a prompt to look for the example closest to your own firm.
Questions to ask a vendor before using a tool on client matters
- Where is our data processed and stored, and for how long?
- Is our content used to train or improve any model?
- Who at the provider can see our content, and under what controls?
- Can we limit access to named users and keep a record of use?
Limits of this article
This summary rests on a single official page. It has not set out the detailed content of the guide, the full conduct rules or how a disciplinary body would view a particular case. If you are drafting a policy for your firm, read the guide in full, check the current rules with the Law Society or the regulator, and take advice where the facts are difficult.
Sources
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
Singapore: the Workplace Fairness Act and what employers must do
Singapore's Workplace Fairness Act has been passed but is not yet in force. This guide explains the protected characteristics, employer obligations and the dispute resolution process, with a plain account of what is not yet law.
6 minute readGenerative AI guidance for lawyers: how six markets compare
How to find what regulators, courts and bar bodies have said in six markets, and the common themes across them.
5 minute readSingapore: the courts' guide on generative AI tools
The Singapore Courts have a guide for court users on generative AI. Read it, then use a checklist.
4 minute read