The Fair Work Commission is taking a closer look at claims written with the help of artificial intelligence (AI). If you're a small business owner facing an unfair dismissal claim, it's possible that the claim wasn't written by a lawyer, but rather by a tool like ChatGPT.
The Fair Work Commission has confirmed that from 20 October 2026, anyone using generative AI to prepare an application or other document for a Commission case will have to say so. They'll also need to check the material is accurate and relevant, and confirm that any witness evidence reflects what the person actually knows, in their own words, not something an AI tool has generated on their behalf.
The Commission published the final Guidance Note on 24 August 2026, alongside a statement from President Justice Adam Hatcher and a commissioned research report into how applicants are actually using AI in practice. It isn't a minor administrative update. The Commission's own figures show its total workload has grown by more than 70% in three years, and its Executive Director has told a Senate committee that somewhere between 40 and 50% of people applying to the Commission say they used AI to help prepare their case.
Why the Commission moved
The pattern the Commission was responding to is one anyone in HR or small business ownership will recognise: dense, professional-looking submissions arriving at volume, some accurate, some not, often from people representing themselves who previously might not have lodged a claim at all. The Commission's own research suggested AI tools were doing three things at once: giving people who weren't eligible to bring a claim confidence to try anyway, lowering the practical barriers that used to stop people from lodging in the first place, and reducing uncertainty for those genuinely unsure of their prospects.
None of that is inherently a problem. Access to information about your rights isn't something to discourage. The issue the Commission is trying to address is what happens when that information is wrong, invented, or applied to the wrong situation, and nobody catches it before it's filed.
What it cost one worker to get this wrong
The clearest illustration of that risk arrived just before the guidance note itself. In a decision handed down on 19 August 2026, the Commission ordered a former Aldi employee, Sadnan Khan, to pay $1,230 towards Aldi's legal costs after pursuing an unfair dismissal claim that had no real prospect of success.
The claim turned on a single date. To bring an unfair dismissal claim, an employee generally needs six months of service, counted from the day they're notified of their dismissal, not the day it takes effect. Khan had started on 29 September 2025 and was told he was being dismissed on 26 March 2026, three days short of the six-month mark. His AI-drafted submissions argued the wrong date, treating his employment as having run into April.
The Commission raised the problem early and gave him more than one chance to reconsider. Commission staff flagged the jurisdictional issue within two days of filing. Days before the hearing, the deciding member wrote to him twice, quoting the relevant law in plain English and warning in one email that "there is a very strong likelihood that your case will not win." He continued anyway, filing further AI-drafted submissions on the same point, and the claim was dismissed at hearing. The Commission's later reasoning noted it was Khan's own responsibility to properly assess his eligibility and to ensure what he submitted addressed the concerns the Commission had raised, regardless of how he arrived at that submission.
Costs orders in unfair dismissal matters are unusual. The Commission is deliberately structured so each side normally covers its own costs, win or lose, precisely so people aren't put off from representing themselves. This one crossed that line because the claim continued after repeated, specific warnings.
Laurence McLean, Director of Operations at workplace relations firm Peninsula Australia, says the new rules reinforce a principle that's easy to lose sight of as AI use spreads through legal and quasi-legal processes: "AI can be a useful tool, but responsibility, judgement and accountability remain with people."
Where AI use cut the other way
A very different case from earlier the same month shows the same tools don't only cause problems for employers. On 12 August 2026, the Commission ruled in Baker v Macquarie University that a casual academic in the university's School of Computing should be treated as a part-time employee, after he'd used AI tools to help prepare his submissions.
The case turned on the "employee choice" provisions added to the Fair Work Act in 2024, which let a genuinely casual employee ask to move to permanent status once their work stops meeting the legal definition of casual. Baker had taught a core computing subject across consecutive semesters, was scheduled to teach it again, and gave the university written notice in November 2025 that he believed his role had stopped being casual. The university declined the notice, and the dispute went to arbitration.
The Commission worked through the statutory test: whether the university could realistically decline to keep offering him the work, whether the work would keep being available, whether permanent staff did the same job, and whether there was a regular pattern to it. Most of those factors pointed toward permanent status, and the Commission found he had not been a genuine casual when he gave his notice.
It's worth being precise about what AI actually did here. This wasn't a case of an AI tool spotting something a lawyer would have missed. The legal test was a fairly conventional application of provisions that had been in the Act for two years. What AI appears to have done is help a self-represented academic identify that he had a case worth bringing, and put it in front of the Commission in a form it could act on. That's a meaningfully different, and more mundane, story than "AI wins landmark ruling," but it's arguably the more important one for a small business owner to understand: the tools are lowering the cost of finding out you might have a claim, not just the cost of writing a bad one.
What this means if you run a business without an in-house legal team
The two cases point in the same direction rather than opposite ones. Whether a claim was AI-assisted or not, the substance still has to hold up: the dates, the eligibility, the facts. From 20 October, you'll at least know when AI was involved on the other side, since disclosure becomes mandatory. That doesn't tell you whether the underlying claim is strong or weak, and it isn't a reason to assume either.
For a small business owner facing an unfair dismissal or general protections claim, the practical takeaway isn't that AI-drafted claims can now be dismissed more easily. It's closer to the opposite: employees now have easier access to a rough map of their rights before they ever contact you, and some of those maps will be accurate. Reviewing your own processes around dismissals, casual employment and record-keeping before a dispute arises is a more useful response than hoping the other side's paperwork falls apart.
This article draws on published Fair Work Commission materials, including the Guidance Note, the President's Statement of 24 August 2026, and the published decisions in the Khan v Aldi and Baker v Macquarie University matters. It is general information, not legal advice.