Usage of generative artificial intelligence by lawyers

“A lawyer without books would be like a workman without tools” was a famous quote by the third President of the United States of America in the early 19^th Century. As time stops for no one, technology has advanced to the point that, in the 21st century, Generative Artificial Intelligence is becoming a common tool for most people, regardless of their field of work.   

One field where Generative Artificial Intelligence (Gen AI) is rapidly emerging as a tool that opens new opportunities is legal services. Most lawyers in New Zealand and overseas are using and investing in this technology to enhance their service offering to clients.

Guidelines for use of Generative Artificial Intelligence (GenAI) in Courts & Tribunals

The New Zealand Judiciary has issued guidelines to assist lawyers who represents clients in court and tribunal proceedings on using GenAI chatbots such as ChatGPT, Bing Chat, or Google Bard. The guidelines were issued on 7 December 2023 and were developed to assist lawyers who represent clients in court and tribunal proceedings.  The guidelines remind lawyers to be cautious and, as officers of the court, not to mislead the court; they must take reasonable steps to ensure the accuracy of information, including legal citations, provided to the court.

What are the consequences if materials submitted are not accurate?

At the time of writing, there are no reported cases in New Zealand in which a lawyer was held liable by the New Zealand Lawyers and Conveyancers Disciplinary Tribunal for using GenAI in preparing and filing submissions with a court and relying on case law and/or research which was inaccurate and misleading and succinctly referred to the courts as “hallucinations”.   

Australia

In the case of Handa & Mallick [2024] FedCFamC2F 957, the Federal Circuit and Family Court of Australia, a Solicitor had submitted a single-page list of authorities to the Judge. Upon returning to the chambers, neither the Judge nor her associate was able to locate the cases identified in that list. The Judge’s associate had asked the solicitor to provide copies of the authorities referred to in the list and was unable to do so.

When the matter resumed, the Judge asked whether the list of authorities had been provided using artificial intelligence. The Solicitor informed that the list had been prepared from LEAP, a legal software package, as the Solicitor understood.  The Judge acknowledged her concerns about the veracity of the information provided in the list of authorities and concerns about the competency and ethics of the practitioner. The practitioner was ordered to make submissions identifying any reason why he ought not to be referred to the Victorian Legal Services Board and Commissioner. A personal costs order against the practitioner was also foreshadowed.

In the case of Dayal [2024] FedCFaC2F 1166, the Federal Circuit and Family Court of Australia on 27 August 2024 had referred the conduct of a solicitor to the Office of the Victorian Legal Services Board and Commissioner. The Solicitor in question had tendered to the court a list and summary of legal authorities that do not exist. The Solicitor confirmed to the court that he prepared the list and summary using an artificial intelligence tool incorporated in the legal practice management software he subscribes to. The Solicitor acknowledged that he did not verify the accuracy of the information the research tool generated before submitting it to the court.

On 2 September 2025, the Victorian Legal Services Board and Commissioner issued a statement on the “Mr Dayal” matter, which varied Mr Dayal’s practising certificate, resulting in him being;

1. No longer being entitled to practice as a principal lawyer
2. No longer being authorised to handle trust money
3. No longer operating his own law practice
4. Only practising as an employee solicitor
5. Undertaking supervised legal practice for a period of two years
6. Both the practitioner and his supervisor reporting to the Victorian Legal Services Board on a quarterly basis during the period of supervised legal practice.

Potential harms caused by filing hallucinated materials with the court

The USA District Court in the case of Mata v Avianca, Inc, 678 F.Supp.3d 443 (S.D.N.Y. 2023) drew worldwide attention to the risk of relying on generative AI for litigation research without independent verification.  The harms identified by the Court are as follows;

1. The opposing party wastes time and money in exposing the deception
2. The Court’s time is taken from other important endeavours
3. Client may be deprived of arguments based on authenticated judicial precedents
4. There is potential harm to the reputation of judges and courts whose names are falsely invoked as authors of the bogus opinions and to the reputation of a party attributed with fictional conduct
5. It promotes cynicism about legal professionals and the judicial system
6. Future litigants may be tempted to defy a judicial ruling by disingenuously claiming doubt about its authenticity.

Important notes to be taken when using Artificial Intelligence tools for research

Check for factual accuracy – ensure that the case citations are correct and read through the actual case to check its authenticity. This applies to every part of an AI-generated document, including facts, dates, names, figures, quotations and legal reasoning and not only the citations.

Cross-check sources – check every case citation, statute, regulation and quotation against the verified legal databases, for example NZLII and/or the official New Zealand legislation website. Read the actual case and/or provision and confirm that cited cases, statutes and regulations actually exist and are current.

Confirm the correct jurisdiction – General AI frequently blends jurisdictions. Ensure the content reflects the correct jurisdiction.

Ensure ethical compliance – don’t enter any information that is confidential or proprietary, or that would reveal the names of clients or other information that could waive privilege.

Think it through and prompt properly – AI-generated information or documents must be individually thought out and properly prompted, setting out clearly what is required before anything is generated.

Require final human sign-off – every piece of AI-generated legal work must be reviewed and approved by a lawyer before release. No exceptions.

Article written by Ashneel Nand