From 20 October 2026, generative AI can no longer help quietly in cases before Australia's Fair Work Commission (below, the Commission). Anyone who has used tools such as ChatGPT, Claude, Copilot or Gemini to draft, rewrite or organise an application, a submission, a witness statement or any other document to be lodged must proactively say that they used it and how, check the document themselves, and state in the document that those checks were completed. The requirements are set out in a statement by the Commission's President and in the Guidance note: Use of generative artificial intelligence in Commission cases, and they apply to all Commission matters from 20 October 2026. This governs Australia's workplace dispute process only — not every Australian court, and certainly not the whole world.
If you used it, say so — and say where
The first requirement in the guidance note is disclosure: if generative AI was used to create, write, edit or otherwise prepare a document, the Commission must be told, with an explanation of exactly how it was used. The Commission's own examples include asking AI to do research and give advice, organise facts and evidence, transcribe voice into text, translate, write content or edit content. From 20 October, some application and response forms will ask directly whether generative AI was used; for documents whose form has no such question, a note at the beginning of the document must state what the AI was asked to do. A bare line saying "this document was AI-assisted" is not enough — the Commission wants the manner of use itself.
The second requirement is verification, and the responsibility cannot be outsourced to the tool. The person lodging the document must personally check that the information is accurate, relevant to the case and supported by real evidence and real legal authorities, and then state in the document that these checks were done. The reason is stated plainly: generative AI can produce incorrect facts, fabricated legal cases, misleading information and outdated law, and giving false or misleading information to the Commission is a serious offence. In other words, if the AI invents a case that does not exist, the person who signs and lodges the document carries the consequences — the tool vendor will not carry them instead. The Commission also states clearly that generative AI is not a substitute for legal advice.
Witness statements and representatives face an extra layer
Witness statements and declarations carry an additional set of requirements: the witness or declarant must confirm that the document is based on their own knowledge, reflects their own words, and is true to the best of their knowledge. This targets the most dangerous use case — polishing a witness's scattered account into fluent testimony while quietly adding details the witness never said and may never have experienced. Once a statement stops being the witness's own words, half of its credibility is already gone.
The bar is higher again for lawyers and other professional representatives: any document prepared using generative AI that refers to case law must include hyperlinks to that case law, so the Commission and the other party can check the original text directly. The direction of travel matches the other side of the Pacific — California's SB 574, which requires lawyers to personally verify AI citations and disclose AI use to the court, likewise pins the verification duty on the practitioner. The difference is that California's rules are state legislation, while Australia's are a procedural guidance note from a single tribunal, binding only on parties and representatives in Commission proceedings.
The cost of ignoring the rules comes in three forms
This guidance is not a polite suggestion. Documents that do not follow the requirements may be given less weight or disregarded altogether; the Commission may order the non-compliant party to pay the other party's costs; and at the application level, a case can be dismissed. All three bite hardest on self-represented parties — the very people who often rely on generative AI to write their applications at all. The guidance does not ban them from using it, but it turns "admit you used it, and check what it wrote" into a procedural threshold: fail it, and the work may be wasted.
One more checkpoint is easy to overlook: privacy. The Commission warns users not to enter personal information about themselves or others, or confidential case information, into public generative AI tools, because what is entered may not stay private or secure. Workplace dispute files are full of exactly this material — pay records, dismissal histories, health information, internal emails. Redacting before use, checking statutory text against Australia's Federal Register of Legislation, and confirming that any cited case actually exists are the most basic pre-lodgement checks.
The same applies to businesses with employees in Australia that may face Commission cases such as unfair dismissal claims: having staff or outside advisers prepare response documents with AI is not prohibited, but disclosure, verification and confidentiality need to be built into the process in advance — not discovered at the Commission's door, when nobody can say which paragraphs the AI wrote or who checked them.
