The FWC recently issued draft guidelines to manage the surge in AI-generated applications, signalling a new landscape of digital compliance for HR. From curbing ‘AI slop’ to maintaining evidentiary integrity, here’s what you need to know to stay ahead of the curve.
Following claims that the Fair Work Commission (FWC) is set to experience a 70 per cent increase in workload over three years due to AI-generated claims from applicants and respondents, FWC President Justice Adam Hatcher recently released a set of draft guidelines on the use of generative AI in Commission cases.
Media reports suggest many of these AI-generated applications have been “baseless claims” that have “pushed the tribunal to breaking point, delaying rulings, compromising major wage cases and making the Fair Work Commission a terrible example of work-life balance”.
“They are looking at both the sources of those claims and also methods of dealing with those claims more efficiently and effectively. So, hopefully, the implementation of these guidelines will either mean fewer claims or, at the very least, claims that are better written,” says Michael Byrnes, Partner at law firm Swaab.
“A significant issue with AI is its tendency to produce voluminous documents filled with irrelevant generalities that often fail to address the specific facts of a case. On occasion, these AI-generated materials even assert false legal propositions or entirely fabricated ‘factual’ scenarios.”
What could this guidance look like in practice?
The draft guidance includes three distinct requirements:
- Documents are to clearly state when generative AI has been used.
- Litigants need to check the document and ensure all details in the document are correct and relevant to the case, and to state in the document that this has been done.
Requirement 2 also lists certain details (such as references to facts and case law) that must be included in the checking of the document. - In addition to requirements 1 and 2, if the document is a witness statement or declaration, it requires the witness/declarant to check the document and ensure it is based on their own knowledge and is true to the best of their knowledge. The witness/declarant must also declare this in the statement/declaration.
“[These guidelines] are highly likely to closely reflect what will end up being implemented,” says Byrnes.
He notes that the first proposed requirement is quite self-explanatory; those submitting or responding to a claim need to disclose when AI is used.
“Without this discipline, HR and People & Culture teams will find themselves tied up in knots responding to ‘AI slop’ that has no basis in law and requires no formal response.” – Michael Byrnes, Partner, Swaab.
“There is a clear distinction between using AI as an aid – such as for grammar checks or transcription services – and using it to generate the substance of an application or response,” he says.
Tools like grammar checks are unlikely to be problematic, he says, as they don’t compromise the integrity of the content.
“They simply ensure the technical accuracy of the writing, confirming that commas are in the right place and that the party or practitioner has avoided malapropisms or similar errors.”
He suggests that AI transcriptions, however, may require disclosure.
“An AI transcription service is a useful tool for recording what was said in a meeting, but it is not without its flaws. You should never tender an AI-generated transcript into evidence without clearly disclosing its source and acknowledging its potential inaccuracies.
“If a transcript is vital to a matter, it must be verified by a human. Much like any draft, the version produced by AI requires a thorough review to ensure it is an accurate reflection of the conversation or meeting.”
Regarding the second proposed requirement, he says: “The responsibility falls on [applicants or respondents] to verify both the legal propositions and the factual assertions within an application or response.
“Ultimately, it is about embedding an element of integrity into any content produced, whether wholly or in part, by AI.”
He says legal practitioners could face additional requirements, which are also being considered for broader applications from employers.
“The proposal currently under discussion is that any cited case law must include hyperlinks. This ensures that the cases can be easily verified as authentic and existent.=
“This measure ensures that parties do not cite false or ‘hallucinated’ cases generated by AI. It also allows the cited authorities to be verified, ensuring they truly support the legal propositions asserted within a Fair Work Commission application, response or submission.”
Regarding the production of witness statements, Byrnes advises against any use of AI at all.
“Witness statements are intended to be a personal recollection of facts – the things you saw, heard and experienced. While people often want their statements to sound eloquent or possess a certain literary quality, that is not their purpose,” he says.
“A statement must be readable and coherent, but above all, it must be your evidence in your own words. AI is antithetical to this process; it applies a ‘machine gloss’ to your voice and, more fundamentally, can distort your unique experience. Using AI is inherently inconsistent with the necessity for accurate, authentic, credible witness statements.”
It could also introduce unnecessary risk if a case makes its way into the Commission.
“It is only a matter of time before witnesses who use AI to draft their statements come unstuck during cross-examination. If the use of AI is not discovered before the matter reaches court, it almost certainly will be under robust questioning,” says Byrnes.
“When AI-generated content is scrutinised by an opposing lawyer or advocate, the witness’s credibility is likely to be impeached – and rightly so.”
A reduced barrier of entry for claims
While the intention of these guidelines is to stem the flow of claims coming through the Commission, it’s still highly probable that it will experience an uptick now that the financial and time constraints of producing a claim have been reduced due to AI.
“HR teams must be properly trained in Fair Work Commission processes, specifically in how to draft responses if they aren’t engaging external counsel,” says Byrnes.
This extends beyond formal proceedings to internal matters like consultation, misconduct investigations and performance management, he says.
“Employers are increasingly being confronted with unreasonably voluminous documents produced or supplemented by AI.
“In these instances, employers should not feel compelled to address every single contention. Instead, they should distill the document down to the essential issues and respond only to what is relevant.
“There is a tendency for employers to treat these documents like a debate or a point-scoring exercise. If an employee submits a 30-page AI-generated document containing 355 points, the employer often feels compelled to counter with a point-by-point rebuttal – even when 90 per cent of those points are irrelevant or lack legitimacy.”
“Using AI is inherently inconsistent with the necessity for accurate, authentic witness statements.” – Michael Byrnes, Partner, Swaab.
In this context, less is more, he says.
“Employers should distill what is truly important and have no compunction about informing an employee that they will not be traversing every matter raised. Without this discipline, HR and people and culture teams will find themselves tied up in knots responding to ‘AI slop’ that has no basis in law and that actually requires no formal response.
“Where there is a reasonable basis to do so, HR practitioners should not hesitate to call out its use. You can adopt a process similar to that of the Fair Work Commission by directly asking an employee if AI was used to produce their documentation. If the answer is yes, the follow-up is critical: what exactly do they want HR to focus on within that document?”
When preparing a response to an employee claim to the FWC, he reiterates that, again, less is more.
“In an unfair dismissal matter, for example, the focus should be on the valid reason for termination, whether procedural fairness was afforded and the other relevant factors, as set out in the legislation, the FWC will consider in making a determination.
“It is essential to filter out the ‘white noise’ or AI-generated ‘sludge’ and concentrate on your core position. This approach aligns with the expectations of the Fair Work Commission. As conciliators and members manage increasing caseloads with tighter timeframes, they do not want to be confronted with 50-page documents from both sides.
“What they require is a concise, disciplined statement of the high-level position, providing just enough information to perform their primary function: conciliating the matter.
“By keeping your response relatively brief and high-level, you are providing exactly what Fair Work Commission conciliators need to move the matter forward.”
Other considerations
There is a significant risk when employees – or even HR practitioners themselves – upload sensitive company information into open-source AI platforms to build or respond to a claim.
By inputting confidential data into these tools, parties risk compromising proprietary information and breaching privacy obligations, potentially creating a secondary legal and security crisis.
“Employees need to be reminded of their obligations in relation to company confidential information,” says Byrnes. “They can’t be inputting that information into open AI models.
“The use of AI in this space is not going away. HR teams must be prepared to manage its influence across all spheres – from Fair Work Commission claims to its use by employees in consultation processes, performance management and misconduct investigations.“
More on AI?
In AHRI’s new podcast, ER/IR Unpacked, host Jonathon Woolfrey FCPHR speaks with employment lawyer Natalie Gaspar on the legal risks of AI in the workplace. Listen to the full episode here or watch the vodcast below.
Want to learn more about using generative AI to enhance HR outcomes? AHRI’s GenAI Integration Essentials course covers the fundamentals of AI, including how to apply it to various HR tasks, be strategic in positioning its role in the HR value chain and address privacy and fairness concerns.
