Law firms are embracing AI: Governing it will be harder

Source: Deeptech Times
Article Date: 23 Sep 2026
Author: Clement Teo

The next phase of adoption will test whether legal organisations can turn AI use into controlled, accountable infrastructure. The real issue is not whether the sector adopts AI. It is how safely and clearly it governs that adoption.

This article was first published on 15 September 2026 in Deeptech Times.

SLW obtained permission to reproduce the article to give the legal community a broader view of legal reports for various news syndicates.

AI is moving into the legal sector fast. In Singapore, the Singapore Academy of Law (SAL) has already rebuilt its flagship research platform, LawNet, around AI. Across APAC, most lawyers now use AI in their daily work.

The shift is changing how legal work gets organised. Research, drafting and review increasingly sit inside connected digital workflows, letting lawyers move faster through large volumes of material.

It also raises harder questions. Lawyers need to know where an answer came from: provenance is central to trust. They need to protect confidential material. And they need to decide when a human must check or stop an AI system before it acts.

Singapore’s drive to lead legal innovation

Singapore is positioning itself as a legal technology hub. In September 2026, Law.com announced a multi-year partnership with SAL to launch LegalWeek Asia, extending the world’s largest legal technology conference into the APAC region.

Building on SAL’s existing TechLaw.Fest event, the inaugural LegalWeek Asia is scheduled for September 28-29, 2027 at the Sands Expo and Convention Centre in Singapore, with organisers expecting more than 4,000 attendees.

The more immediate change happened within LawNet itself. At TechLaw.Fest in September 2025, SAL launched LawNet 4.0, adding an AI-powered search tool built with Singapore’s Infocomm Media Development Authority (IMDA), a new legislation module linking statutes to case law (starting with the Companies Act), and an upgraded Academy Library for legal texts.

This points to a clear aim. LawNet is being shaped as a research environment, not just a document store. Lawyers still need primary sources and independent legal judgment, and they now have quicker ways to move across statutes, cases and commentary.

Adoption across the region is already widespread. The ALITA State of Legal Innovation in APAC (SOLIA) Report 2025, based on a survey of 287 legal professionals across the region, found that nearly 90 per cent of respondents already use AI at work.

Most everyday use remains practical. Lawyers use general-purpose AI to draft emails, summarise documents, translate text and prepare notes. These tasks save time without necessarily sitting at the centre of legal judgment.

Specialist legal AI sits closer to the work lawyers are trained to do. It supports document review, document analysis, research and drafting. These uses demand stronger controls, since the output can shape advice, negotiation or litigation strategy.

For a legal team, the real test is trust. A system producing a fluent answer is not enough. The answer must trace back to reliable legal material.

Time is the strongest driver. Much legal work depends on finding, reading, comparing and summarising information, a category AI handles well. Client demand adds further pressure. In-house teams want faster answers for the business, and law firms want to improve efficiency without simply adding junior hours. Vendors, in turn, see an opportunity to bring research and drafting into a single interface.

That pressure is likely to grow. Once lawyers see AI shorten routine tasks, they start expecting the same speed across more complex workflows. That is where the governance challenge begins.

Tackling accuracy, secrecy and accountability

Accuracy is the first problem. Legal AI cannot simply sound convincing; it must cite authenticated statutes, judgments or trusted commentary. A false answer can affect advice, rights and professional duty.

Confidentiality is the second problem. Lawyers handle privileged material, commercial strategy, personal data and court-sensitive information. They need clear rules on model choice, data retention, access rights and vendor checks.

Accountability is the third problem. An AI tool may draft a clause, rank authorities or suggest a litigation route, and the lawyer and the organisation remain responsible for the result. Governance has to sit inside daily work, not alongside it.

Why agentic AI raises the stakes

Agentic AI can plan work, call tools and carry out a sequence of steps on its own. In law, that could support research bundles, filing alerts, workflow triggers and corporate secretarial processes.

This makes oversight more important. A system that only answers a question creates one type of risk. A system that acts across tools creates a wider one.

The Singapore Computer Society’s Agentic AI Ethics and Governance Body of Knowledge, Version 2.1, treats agentic AI as both a technology issue and a governance issue. Its central question is simple: how much agency should a system have, what can it access and when must a human step in?

The next phase of legal AI will reward organisations that treat adoption as infrastructure: trusted legal content, human review for consequential work, measurable return on investment and clear governance controls.

The strongest teams will define tasks carefully. Some work can be automated. Some work should be augmented. Some work should stay firmly under human judgment.

That may be the biggest change ahead. AI will not remove the need for lawyers to reason, check and decide. It will shift where those skills get applied.