Experts say guidelines for AI use in New Zealand's courts need to be updated, but the courts say a cautious approach is still needed.

Generative artificial intelligence (GenAI) use in New Zealand's legal system is continuing to increase, a fact the Office of the Chief Justice has acknowledged.

But the judiciary's guidelines for AI use were last updated in 2023.

Matt Bartlett, a lecturer at the University of Auckland's Law School, said other judiciaries around the world were updating their guidance far more frequently than New Zealand.

"They're out of step with the capabilities, the risk profile, and frankly the practice overseas," he said.

Dr Joshua Yuvaraj, a senior lecturer at the University of Auckland's Law School, said the technology was evolving so quickly that the guidelines should be reviewed regularly to ensure they remained fit for purpose.

"Three years is a very long time in the technology space," he said.

Technology, he said, had moved beyond just generative AI technologies to agentic AI, where AI systems could automate tasks such as organising workflows, responding to emails and other administrative work.

He said these developments warranted some attention in the court's guidelines. He also wanted to see continual emphasis on the privacy and confidentiality risks associated with using AI tools.

Auckland barrister Josh McBride said while the guidelines were now a few years old, they had largely stood the test of time.

"I've looked at them recently, and in fairness, they do a pretty good job," he said.

"The problem with getting prescriptive and being detailed about what tools are okay and which aren't, and what uses are okay and which aren't, is that you are having to literally update these every few weeks, which is a very, very difficult task.

"I think all that the courts can do, and I think they're doing this in New Zealand really well, is to say, 'proceed with extreme caution if you're using these tools in this context. You are responsible for the material that goes into the court. You can't blame a robot if it goes wrong, and we expect you to comply with all of the ethical obligations'."

More transparency needed

Typically AI use did not need to be disclosed, and Bartlett said that made it difficult to understand the extent to which AI was already influencing the justice system.

He said courts overseas were already experimenting with AI, and it would be naïve to assume New Zealand judges would not use similar tools.

"The danger with an AI is that it undermines what we currently have from judges, which is them explaining their reasoning.

"They're writing their judgments. They are explaining how they arrived at their legal conclusions. If you have an AI that is writing the judgment, you're undermining the explanatory process of how a judge arrived at their decision, which is just core to how the whole common law, our whole constitutional structure operates."

Bartlett questioned whether overseas AI models could adequately account for New Zealand's unique legal and constitutional context.

"They will place no particular emphasis on New Zealand-specific constitutional factors."

He noted the judiciary's own guidance already warned that GenAI chatbots did not adequately account for New Zealand's cultural context or the specific cultural values and practices of Māori and Pasifika.

Yuvaraj said if AI use wasn't disclosed then people didn't know how much it had been used and therefore whether it had any potential impact on the outcome of a case.

"We want to make sure that the judge has read everything that's relevant rather than just outsourcing that, because then we can't be sure that the judge has administered justice fully."

McBride meanwhile did not believe AI use should necessarily have to be disclosed.

However he said he had seen instances where litigants armed with AI tools were producing disproportionate volumes of AI generated material for what the case entailed.

He said there was a benefit in clearer rules being developed around the volume of AI-generated material people could file, along with clearer warnings that litigants were responsible for verifying anything AI produced before submitting it to the court.

Courts taking cautious approach

In response to questions from RNZ a spokesperson from the Office of the Chief Justice said while the judiciary recognised the opportunities created by AI to support access to justice and the timely, efficient and proportionate administration of justice, it was taking a cautious approach to the exploration and adoption of Generative AI (Gen AI) to support the work of the courts.

"At present, a limited number of judges and their support staff are trialling some Gen AI tools overseen by the Judicial AI Advisory Group. The trials start from the premise that a judgment is a reasoned human decision made by a judge drawing on the evidence, submissions, pleadings, charges, and applicable law. Judgment writing is a fundamental responsibility of the judge - that responsibility cannot be delegated to other people, or to AI."

The spokesperson said the guiding principles for the trials were that Gen AI tools must not be used for judicial reasoning or decision-making, and must not be used in a way that replaces, compromises, or obscures the essential human task of judging.

"The trials are testing responsible use cases of AI, and are being used to refine the existing Guidelines for Judicial Use of AI - these were issued in 2023 and need updating.

"The objective is to ensure that guidelines and guidance are fit for purpose and support the safe and appropriate use of Gen AI in the courts, and provide public confidence regarding the extent of use of AI by judges."