AI Regulation Tracker  /  Professional-body guidance

The Law Society of NSW Maps Solicitor AI Duties Onto Existing Conduct Rules

The Law Society of New South Wales has published A Solicitor's Guide to Responsible Use of Artificial Intelligence, updated January 2026. It is guidance, not a new rule. It does not create a duty. What it does is spell out how the existing, binding Solicitor Conduct Rules already apply when a solicitor uses generative AI: check every output, keep client information out of public tools, and never let AI mislead the court.

By Anthony Guerriero  ·  The Leveraged Years Regulatory Research  ·  Published July 17, 2026  ·  Last updated July 17, 2026

The Law Society of New South Wales is the professional body for solicitors in Australia's largest and busiest legal jurisdiction. In January 2026 it published an updated version of a guide with a plain title: A Solicitor's Guide to Responsible Use of Artificial Intelligence. It runs to a short, practical document, and it is worth reading closely precisely because it does not try to do anything dramatic. It does not announce a new rule. It takes the duties solicitors already owe and walks them through, one at a time, against the reality of generative AI.

The guide states its own logic up front. It conveys, in its words, a simple principle that applies to any new technology in legal practice: understand the tool you are using and know how it can assist or damage your practice, and know what your professional and ethical obligations are and how they apply to that tool. That framing tells you what this document is. It is not AI policy. It is professional-conduct policy applied to AI.

The five rules it maps

The heart of the guide is a section that runs through five of the binding Solicitor Conduct Rules and explains how each bears on AI use.

On competence, Rule 4, the guide is direct. Solicitors are responsible for the accuracy of any advice they give, regardless of where they did their research, and generative AI can produce fabricated material including fictitious case law. On accuracy generally the guide puts it bluntly: because a generative AI tool may be quick and easy, "solicitors must therefore always thoroughly check its outputs to avoid compromising their integrity." The reason it gives is the one every practitioner should keep pinned to the wall. As the guide says, "Humans are accountable, but generative AI is not."

On confidentiality, Rule 9, the guide draws the sharpest line in the document. Generative AI takes what you enter, learns from it, and can surface it to other users. So, in the guide's words, "any placing of client confidential information into a publicly available generative AI system is akin to putting it in the public domain. This is likely to be a breach of confidentiality and clients may lose privilege." That is not a maybe. Pasting a client's confidential material into a public chatbot is treated as disclosure.

On independence, Rule 17, the point is that solicitors must exercise their own judgment and are not a mouthpiece, whether the views come from a client or from an AI system. On the duty to the court, Rule 19, the guide is equally firm: solicitors must ensure they do not mislead or deceive the court even inadvertently, and the validity of any material presented to the court needs to be tested by solicitors, whether or not that material was produced by generative AI. And on supervision, Rule 37, it warns that a supervising solicitor needs to be particularly cognisant of AI risk, and that meaningful supervision requires critical evaluation of the accuracy and completeness of the tool's outputs, something a junior may not have the experience to do alone.

Real cases, not hypotheticals

What gives the guide weight is that it does not argue from imagined risk. It cites actual matters where lawyers got burned. It references an Australian case in which a practitioner had his practising certificate varied by the Victorian Legal Services Board after presenting hallucinated authorities, and a native-title matter, Murray v Victoria, in which a junior solicitor working remotely prepared a document with footnotes citing reports and papers that were mostly non-existent or incorrectly cited, with the court finding both that she did not take sufficient care and that her supervisor failed to ensure the work was adequately supervised.

Notably for a US audience, the guide also cites Mata v Avianca, the Southern District of New York matter where a lawyer relied on ChatGPT to confirm cases that turned out to be fictitious. An Australian professional body reaching for a US case to make its point is a useful signal in itself: this is a common-law problem, and the professional duty is converging across borders.

What this is, and what it is not

I want to be precise, because it is easy to over-read a regulator's name on a document. This is a guide. It is guidance issued by the professional body, and it says so in its own framing as a guideline that reminds practitioners of obligations that already exist. It does not create a new duty, a filing requirement, or a disclosure mandate that was not already there.

What it interprets, though, is binding. The Legal Profession Uniform Law Australian Solicitors' Conduct Rules 2015 are enforceable professional rules for NSW solicitors, and the recent cases the guide cites show what happens when they are breached in an AI context: varied practising certificates, adverse court findings, supervision failures on the record. So the correct read is that the guide itself is non-binding, but the duties it describes are real and enforceable, and the guide is telling you how they now apply to a tool you are probably already using.

What this means for US lawyers

The Law Society of NSW is not your regulator, and this guide carries no authority over a US practice. Its direct value is for solicitors in NSW and, given cross-border work, for US lawyers who touch Australian matters or supervise Australia-facing teams. If your practice reaches into that jurisdiction, this is now the professional body's stated expectation, and you should read it the way you would read your own state bar's guidance.

For everyone else, the value is comparative, and it is real. Strip out the rule numbers and the duties are the same ones every US lawyer already owes: competence, confidentiality, candor to the tribunal, and supervision of subordinates. The NSW guide reaches the same conclusions US courts and bar authorities have been reaching, that you must verify every AI output before it goes anywhere near a court, that client information does not belong in public models, and that a supervising lawyer cannot outsource the check to someone who cannot actually evaluate the tool. When an independent common-law regulator lands on the same duties from a standing start, that tells you the AI-ethics standard for lawyers is converging, and it is not moving in a direction that excuses unverified AI work. Treat this as a benchmark for your own AI policy, not as authority you are bound by.

What to do now

Read the actual duties, not the headline. Verify every citation and every factual claim in anything AI touched before it is filed or sent, because the duty to the court and the duty of competence both sit on you, not the tool. Keep confidential client information out of public generative AI systems, and treat any such input as potential disclosure that can cost privilege. If you supervise, make sure the person checking AI output can actually judge whether it is accurate and complete. Write down your firm's AI policy so these expectations are not left to memory. And do not read a professional-body guide as permission to relax. It is the opposite. It is the standard you will be measured against if something goes wrong.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule or professional obligation applies to your situation with qualified professionals in the relevant jurisdiction.