FWC’s guidance note forces applicants to ‘stand behind’ AI

Neil praised the core requirements of disclosure, verification and authenticity that the AI Guidance Note provides, but noted that applicants will likely struggle, specifically those who base their applications on knowledge from AI.

“Verification requires that every reference in a document that has been created using AI, every reference to a decision, to a statutory provision, to an aspect of the evidence, must be verified by the person creating the document, personally checked by them,” he said.

“And I think that requires a degree of knowledge and familiarity with the underlying source material that many applicants will struggle to possess.”

“I think the real problem is will applicants be able to do what the guidance note requires?”

This sentiment outlines the core issue of AI – especially in a legal setting – that while it may open doors and improve accessibility on the ground floor, it is unlikely to achieve favourable judicial outcomes in the long term.

In this vein, Neil said the idea of AI is often more alluring than its practice, adding: “More has been promised than has been delivered.”

VIEW ALL

Speaking to Lawyers Weekly earlier this month, Kingston Reid partner Michael Stutley said success stories – such as the dispute brought by computing academic Gregory David Baker – involved “very sophisticated users of AI which have been proficient in training agents”.

“This is far from the everyday user of a commercially available free version of generative AI,” Stutley said.

Tinashe Makamure, special counsel in Barry Nilsson’s insurance and health team, identified that “GenAI is a double-edged sword”.

“It has, on the one hand, improved accessibility in that more people seem willing to engage with the legal system through the use of GenAI. That being said, there has been a significant increase in frivolous claims with those usually involving voluminous material and placing a strain on the limited resources that courts and tribunals have,” he said.

According to Makamure, this delay in outcomes has a negative effect on accessibility.

Neil explained that the quality of AI remains the core issue for those using and receiving content, not only taking up commission time with a greater amount of applications but also content that features, as it is colloquially known, AI slop.

“AI is marked by sycophancy and hallucinations. It’s designed to tell people what they want to hear. And that’s been a real problem in the legal system generally,” he said.

Accepting that it is too soon to fully gauge the impact of AI in the sector and and FWC’s AI Guidance Note, Neil said there needs to be a more common approach across all courts and tribunals in Australia.

For the time being, Neil surmised that the guidelines only reinforce that “no one can rely entirely on AI to generate the content of documents … they have to apply their own mind to it or at least there must be some human intervention”.

“A person must stand behind what the AI creates,” he said.

As recently covered, legal and HR experts generally praised the introduction of AI regulation within the commission, but questioned how it would operate in practice – especially for those without legal expertise at the centre of application.

Makamure urged employers to acknowledge and prepare for a likely increase in obligations, which may necessitate extra training for HR and legal teams, and a review of internal policies governing AI use.

“Employers must also avoid entering personal or confidential information including names, contact details, health records and information obtained through Fair Work Commission orders in public GenAI tools, as this could result in breaches of privacy obligations or commission orders,” he said.