Losing your job is stressful enough. When hiring a lawyer feels beyond reach, reaching for a chatbot to draft your unfair dismissal claim feels like a reasonable shortcut. The Fair Work Commission (FWC), Australia's national workplace relations tribunal, has now formalised exactly what that shortcut costs you if you get it wrong - and the price can include a bill for the other side's legal fees.

Since 20 October 2026, anyone lodging documents with the FWC must disclose whether generative AI helped prepare them, and explain how. That duty falls on workers and employers alike.

Why the FWC stepped in

The Commission's total workload rose by more than 70% over the three years to June 2026, and FWC President Justice Adam Hatcher has attributed much of that growth to generative AI making it easier - and faster - to file unfair dismissal and general protections applications. Analysis published in The Conversation recorded a total of 44,039 claims lodged between July 2025 and April 2026 alone, a figure that underscores just how accessible the Commission has become for self-represented applicants.

Justice Hatcher's office released the Guidance Note on the use of generative AI on 24 August 2026, giving parties close to two months to prepare before it took effect. Documents lodged before 20 October 2026 were not covered. Everything filed from that date onward is.

The case that set the tone: Khan v Aldi

The Guidance Note did not arrive in a vacuum. It followed Sadnan Khan v Aldi Pty Ltd [2026] FWC 3144, a decision that illustrated exactly what goes wrong when a dismissed worker treats a chatbot as a substitute for legal advice.

Mr Khan started as a store assistant at ALDI on 29 September 2025 and was told of his dismissal on 26 March 2026. That gap left him three days short of the six-month minimum employment period required to bring an unfair dismissal claim under the Fair Work Act 2009 (Cth). His application was drafted with generative AI and was submitted with visible AI drafting notes still embedded in the text.

Deputy President Easton, deciding the case, observed that it was quite possible Mr Khan had adopted AI as a quasi-legal adviser and then submitted the AI-generated outputs as his own material without meaningful review. The Deputy President went further, finding that if Mr Khan had read the AI-generated material carefully and cross-checked it against the FWC's own website and the guidance he had already received from the Commission, he would have recognised that his AI-based arguments were simply incorrect.

The costs consequence was significant. The FWC typically operates on the principle that each party bears its own legal costs. Deputy President Easton departed from that norm and ordered Mr Khan to pay A$1,230 towards ALDI's legal costs under section 400A of the Fair Work Act 2009 (Cth) and Schedule 3.1 of the Fair Work Regulations 2009. ALDI had warned before proceedings continued that it would seek costs. Mr Khan pressed on.

AI as a drafting tool is fine. AI as your legal adviser is not.

The same decision is careful to draw a line rather than a blanket ban. Deputy President Easton's ruling makes clear that using AI to assist in drafting correspondence and even submissions is not, of itself, unreasonable, and that AI in 2026 is a drafting tool that can assist litigants when used properly.

The distinction is between using a tool to organise your thoughts and using it to tell you whether you have a claim. The first is practical. The second is where the risk sits entirely with you.

Tribunal rules also vary. The NSW Industrial Relations Commission takes a stricter position, prohibiting the use of generative AI when preparing witness statements and evidence. Before you draft anything, check which body is hearing your matter and what its current rules allow. The FWC and the NSW IRC are different forums with different standards.

What the FWC rules ask of you

Natasha Jones and Michael Hope, Special Counsels at national law firm Holding Redlich, have published analysis confirming that the Guidance Note covers every party before the Commission - not just dismissed workers, but employers and their representatives as well. In practical terms, the obligations work like this:

• Disclose. State clearly in the document that generative AI was used and describe how - for example, to structure a timeline or improve readability.
• Verify. Every fact, date and legal reference the tool produced is your responsibility. Cross-check each one before you lodge. The Commission holds you accountable for the content, not the software.
• Link the cases you cite. Legal representatives are required to hyperlink every case they rely on. If you are representing yourself, the Commission encourages you to do the same - it is also the fastest way to confirm a case exists and says what the AI claims it says.

Failing to follow these rules carries real consequences. The Commission can give your documents less weight, make an adverse costs order against you, or dismiss your case entirely. There is also a criminal dimension: providing false or misleading information to the FWC may breach section 137.1 of the Criminal Code Act 1995 (Cth). An invented case citation is not a harmless formatting error.

How to use AI on a dismissal claim without it backfiring

If you have been dismissed and plan to use a chatbot to help draft your application, the following steps keep you within the rules and reduce the chance of an avoidable outcome:

1. Check your own eligibility first. Count the exact number of days from your start date to the date you were told about the dismissal, then compare that number with the FWC's eligibility guidance on its website. The Khan case turned on three days.
2. Use AI for structure, not legal judgement. Ask it to put your timeline in order or tighten your language. Do not ask it to assess whether you have a strong case - that determination depends on specific facts and current law, and chatbots get both wrong in ways that look plausible on the page.
3. Read every line before you lodge. Remove any visible prompts, placeholders or drafting scaffolding the AI left behind. Visible AI notes were a central problem in the Khan application.
4. Keep a brief record of how you used AI. A short note to yourself - for example, that you used a chatbot to draft an initial timeline on a particular date - makes your disclosure statement accurate and straightforward to write.
5. Take formal warnings seriously. If the Commission or your former employer raises a problem with your claim, such as a question about your minimum employment period, pause and reassess before continuing. Pressing on with a claim that has no reasonable prospects is precisely what converted Mr Khan's application into a costs liability.

What has not changed for you

The Guidance Note does not remove your right to bring a claim. It does not stop you using AI. What it does is formalise responsibilities that responsible litigants would already follow: disclose the tools you used, check the output, and stand behind what you file.

The rules also cut both ways. Your former employer is bound by exactly the same disclosure and verification obligations if it uses generative AI to draft its response. If something in its documents looks unchecked or factually inaccurate, you have every right to raise that with the Commission.

Running a dismissal claim alongside a job search is the reality for most people in this position. The claim process can take months, and keeping your options open protects your income whatever the outcome. Starting early with job applications is a practical parallel track, not an admission that the claim will fail.

Generative AI can help you write faster and more clearly. It cannot confirm that you were employed long enough to bring a claim, and it will not pay the A$1,230 if it gets the law wrong. Read what it produces, check it against official FWC guidance, and put your name only on what you know to be accurate.