AI tells tenant she should ask for $40,000 – tribunal hands her $80

Source: Radio New Zealand

Tenants using artificial intelligence to help them make applications to the Tenancy Tribunal are creating extra work and backlog, one property manager says.

Property Brokers general manager of property management David Faulkner said there had been a noticeable shift towards AI-written applications that tended to be longer, more complex and sometimes claiming excessive amounts of money.

He said his firm received a notice of hearing for a tenancy dispute just before Christmas. A tenant had claimed that she and her child had their safety compromised when they were renting a rural property with unsafe drinking water.

“There were other issues as well – a claim of retaliatory notice, breaches of quiet enjoyment, and a dryer breaking down, which took a total of four weeks to repair. Her claim was for $40,000.

“What we received was a total of 215 pages, made up of the application, evidence, photographs, and a 101 page written report outlining the claim and breaking down the costs that should be awarded to her. The basis of the claim was that the pH level of the water was at a level that made it dangerous to drink. There were two hearings, one remote and one in person.

“On 2 April, the adjudicator published the Tribunal order. The tenant was awarded a grand total of $80 for the inconvenience of the dryer.”

He said excessively long, seemingly AI-generated claims like this put pressure on staff as well as the owners of the property.

“[That] was probably the first one, and we’re thinking this looks very, very complex. Then you start to see two or three more come through asking for really big amounts, $40,000, $50,00, $60,000 and they’re all set out in the same format. It just becomes quite clear that it’s AI-generated.

“A tenant may have a grievance, they put it together, and, you know, in good old days they’d probably go to somebody like a tenancy union or a tenant advocate, and at least could have a conversation with them, but AI in some cases is just giving them information which is just not being verified. And they’re probably getting very excited about what they think they can get, and they submit it in. And it’s started to cause a few problems.”

He said Tenancy Tribunal adjudicators were having to go through all the pages of evidence and it slowed the process.

Faulkner said some claims did not have merit and others were exorbitant.

Landlords submit the bulk of Tenancy Tribunal applications, commonly for rent arrears. Faulkner said these cases often did not need to be dealt with via a hearing and the tribunal could speed up the rate at which it heard cases by dealing with hem remotely.

Sarina Gibbon, director at Tenancy Advisory, said it was true that cases were being lodged that were more complex and sometimes meritless.

But she said it was probably another stage in the evolution of technology.

“If you look at the overall system in terms of the common challenges we have now which is to make the tribunal pipeline flow more effectively and minimise wait time so genuine applications are not having to wait months and months and months to get a hearing, in that sense having a very primitive grasp of AI is not helping.

“But if you take a step back and consider the overall intent of a tenancy-specific tribunal, which is really set up to promote cheap and expedited justice or dispute resolution … if AI enables more applicants to put forward applications, is that helpful? Yeah, I think it’s helpful. It’s serving justice and giving access to justice to more people.

“I don’t see a problem with AI as such, I think this is a reflection of simply that the technology is just so new and we’re still such naive and primitive users that we’re grasping with what this technology means. So right now what is happening is a lot of people are engaging with AI with very little knowledge of the RTA themselves. So they’re buying into AI hallucination, they’re buying into this tremendous amount of bloat that AI produces.

“I’m a lover of technology, over time I’m confident that AI is going to get us to a point where it’s going to be, it’s going to be a help rather than a hindrance.”

It comes after the Insurance and Financial Services Ombudsman warned last week that people should fact-check the information AI was giving them when they made complaints.

In one example seen by the IFSO Scheme, a Google AI summary suggested that insurance claim decisions are “frequently overturned” when consumers complained, and that “up to 80 or 90 percent of cases can result in success if people persist”.

But Insurance and Financial Services Ombudsman Karen Stevens said that was misleading.

She said complaints that escalated were often complex and stressful, and not always able to be resolved in the consumer’s favour.

She said AI responses could make the process more frustrating for people when reality did not match their expectations.

“While AI can be very useful for general information, it can sometimes oversimplify complex policy wording, miss key exclusions, and rely on overseas information that doesn’t apply in New Zealand. We’ve also seen instances of AI hallucinating previous cases and using them as examples. This can result in it giving incorrect advice,” she said.

“We’ve seen complaints which are 300 pages long. But more words aren’t necessarily better. Clear information about what has gone wrong for someone is much more useful than multiple pages referencing legislation and case law.”

Sign up for Money with Susan Edmunds, a weekly newsletter covering all the things that affect how we make, spend and invest money.

– Published by EveningReport.nz and AsiaPacificReport.nz, see: MIL OSI in partnership with Radio New Zealand