Australia's workplace tribunal says it's getting overwhelmed by requests from people using AI when submitting cases. Its solution? More AI.
An increasing number of cases informed by AI-generated content are being brought to the tribunal, and it's on track to increase its workload by 70% over the last three years, the general manager of Australia's Fair Work Commission, Murray Furlong, said in a report published Friday.
Across the legal world, AI has lowered the barrier to filing a case. Institutions like Australia's Fair Work Commission are asking whether they can fight fire with fire and use AI to manage the rising workload.
Furlong said the rise of generative AI tools has been coupled with more people choosing to represent themselves in cases, as well as budget constraints and resourcing challenges.
"These impacts, taken together, are having a direct effect on the Commission's ability to provide timely, efficient and effective dispute resolution services to the community," Furlong added.
To improve efficiency and cut down costs, Australia's workplace tribunal, which resolves disputes like unfair dismissals, is turning** **to AI solutions.
These include using generative AI to automate case processing, developing an AI tool to assist with pre-approval checks for applications, and an AI voice agent to help triage helpline calls.
It's an issue that's playing out globally. The rising use of AI in US federal civil court cases is placing a greater burden on courts, according to a research paper published by Anand Shah of the Massachusetts Institute of Technology and Joshua Levy of the University of Southern California in May.
One solution Shah and Levy proposed in the paper was relaxing the rules that prohibit federal judges from using AI assistance to increase their productivity levels.
Instances of lawyers filing documents that include fake cases have spooked the legal world. To combat AI hallucinations, some legal tech startups, such as Harvey, are working with legal databases like LexisNexis to limit what information their large language models can access.