On 24 August 2026 the President of the Fair Work Commission (Commission), Justice Hatcher, published a statement together with the final version of the Commission’s Guidance Note, Use of generative artificial intelligence in Commission cases (Guidance Note). From 20 October 2026, anyone who uses generative AI to help prepare a document lodged in a Commission case, whether an applicant, an employer, an HR adviser, a union official or a lawyer, must disclose that use, check the document and say that they have done so. This insight explains why the Commission has acted, what the Guidance Note requires, and what it means in practice for everyone who appears before the Commission.

Why the Commission has acted

The Commission’s workload has grown by more than 70 per cent in three years. By the end of April 2026 it had received 44,039 lodgments for the financial year, just 36 short of the entire 2024–25 total with two months still to run, and the research underpinning the Guidance Note forecasts around 53,600 lodgments for 2025 – 26. The Commission attributes this growth principally to the use of generative AI by potential litigants. The surge is not explained by the labour market, and it has coincided with the widespread appearance of AI-generated language in applications.

To understand what was happening, the Commission commissioned independent research in February 2026, comprising 18 interviews, surveys of 408 applicants and 211 respondents, and consultations with Commission Members, staff and stakeholder groups. The findings, published alongside the President’s statement, include that:

• approximately 40 per cent of surveyed applicants used AI to prepare or manage their case, with use concentrated among self-represented applicants, younger applicants and those in professional and managerial occupations, and twice as common among non-native English speakers;
• ChatGPT was the tool used by around 77 per cent of AI-using applicants, most of them on the free tier;
• 57 per cent of respondents believed the applicant in their case had used AI, while respondents’ own use was more restrained and largely confined to organising responses and supplementing professional advice;
• AI is contributing to lodgment volumes through several mechanisms, including redirecting applicants who are ineligible for unfair dismissal claims into general protections claims, lowering the administrative and emotional barriers to lodging, and generating confidence to proceed regardless of whether the underlying assessment is sound; and
• AI use falls along a spectrum. “AI-assisted” users cross-check outputs and exercise independent judgment. “AI-dependent” users accept outputs at face value, treat AI as a decisive source of guidance on prospects and remedies, and arrive with “confident but fragile positions”. Given AI is “designed to be helpful, agreeable, and responsive to what users want”, sycophantic and hallucinatory outputs inflate expectations, and those most exposed are the users least likely to check.

The Commission published an exposure draft of the Guidance Note in March 2026, received 28 submissions, and finalised it with those submissions and the research in mind. Its response is deliberately not a ban. The Guidance Note recognises that AI can improve efficiency and support access to justice for self-represented parties, while setting the standards that anyone using it must meet.

Who and what the Guidance Note covers

The Guidance Note applies whenever generative AI is used to “create, write, edit, modify or otherwise prepare” an application or any other document lodged in a Commission case. “Document” is defined broadly and includes Commission forms and templates, submissions, chronologies, declarations, witness statements, statements of agreed facts and other evidentiary material. Using AI only to check spelling or grammar, or to format a document, is not caught.

It applies to every participant. The Commission will also modify its forms in all individual rights matters so that they contain an AI disclosure section and a checking declaration. Every person completing a form will be required to complete the AI section and the declaration, whether or not AI was used.

The three requirements

1. Disclose. You must state in the document that generative AI was used and how. The modified forms will offer a list of uses to select from: researching or getting advice about the law, eligibility, outcomes or Commission processes; organising facts, evidence or the timeline; transcribing speech or translating into English; writing content used in the document; modifying or editing content you created; or other uses. In any other document, a statement that “GenAI was used in preparing this document”, together with a description of how, will suffice.
2. Check, and say you have checked. Before lodging, you must check the document and correct it so that every detail is accurate and relevant. In particular, all references to facts and evidence must be correct and the facts and evidence must exist. Every case, statute, textbook or article cited must exist and support the proposition for which it is cited, and every extract or quote must be accurate and come from the stated source. The Guidance Note is explicit that this checking must be done by a person, using sources known to be reliable such as the Commission’s Benchbooks, its decisions database, AustLII, Jade and the Federal Register of Legislation. Asking the same or a different AI tool to check the document will not be sufficient. Hyperlinks to all cited cases are recommended for everyone and required for professional representatives. The document must then state that it has been checked.
3. Keep witness statements and declarations in the witness’s own words. AI may be used to transcribe a person’s speech or translate it into English, but it must not be used to create the factual content of a witness statement or declaration. The statement must be based on the witness’s own knowledge, reflect their own words and be declared true to the best of their knowledge, and it must say so.

Personal and confidential information

The Guidance Note also addresses what goes into the AI tool, not just what comes out. Parties should not give personal information about another person involved in the case, or confidential information about the case, to public AI tools such as ChatGPT, Claude, CoPilot or Gemini, or to any AI tool that may not keep the information secure. The Commission recommends that parties do not give their own personal information to public AI tools either.

“Personal information” includes names, contact details, health information, employment history, financial information, union membership and allegations of misconduct. “Confidential information” includes anything said in settlement discussions at a Commission conference, documents obtained through a Commission order, material subject to a non-publication order, legal or financial advice, and information covered by a confidentiality clause in an employment contract. Giving an AI tool material that is subject to a Commission order restricting publication may breach the order and constitute an offence under the FW Act.

What happens if you do not comply

Non-compliance may result in a document being given less weight or disregarded, a costs order against the party, or dismissal of the case. Disclosing AI use will not itself affect how the merits of a document are assessed, provided the requirements are met. However, failing to disclose it might.

The Guidance Note also reminds parties that knowingly giving false or misleading information to the Commission is an offence under section 137.1 of the Criminal Code, and that giving false or misleading evidence is an offence under section 678 of the Fair Work Act 2009 (Cth) (FW Act). Both carry a maximum penalty of 12 months’ imprisonment.

The costs risk is not theoretical. Days before the Guidance Note was finalised, the Commission ordered an applicant who had relied on AI to commence and continue a hopeless unfair dismissal claim to pay part of his former employer’s costs, a step it described as “very rare”. We examine that decision, Khan v Aldi Pty Ltd [2026] FWC 3144, in our companion insight.

Lawyers and professional representatives

The standard is higher for those who represent others. A “professional representative” includes a lawyer or paid agent, a human resources adviser employed by a party, and an employee or officer of a union, employer organisation, peak council or not-for-profit body representing a party.

Professional representatives who use AI to prepare a document must include hyperlinks to all case law cited in it.

Legal practitioners must in addition comply with their professional and ethical obligations in any use of AI in a Commission case, and a failure to do so may be the subject of a complaint to their legal services regulator.

What it means in practice

For employers and HR teams. In-house HR advisers are expressly professional representatives, so the hyperlinking requirement and the higher standard apply to them. Responses and submissions prepared with AI will need to carry the AI disclosure and the checking declaration, and the person who signs must actually have done the checking.

Witness statements are a particular risk area. A statement that is generated with AI from a manager’s notes will not satisfy the requirement that it reflects the witness’s own knowledge and words.

Employers should also look hard at what is being pasted into public AI tools. An applicant’s personnel file, medical certificate or the contents of a conciliation conference should never go into a consumer chatbot, and where AI is used at all it should be an enterprise tool with appropriate data protections.

For employees and applicants. AI can help with organising the facts and timeline, understanding Commission processes and expressing a case clearly, and the Commission has published its own guidance and suggested prompts on using it well. It should not be relied on to predict whether a claim will succeed or what it is worth. The research found that AI tends to affirm the user’s position and to produce remedy estimates at or beyond the realistic range. Ask the tool to argue against your case, check every citation against the Commission’s Benchbooks, and consider obtaining legal advice before lodging.

For lawyers, agents and unions. The Guidance Note sits on top of existing professional obligations rather than replacing them. Firms and organisations should make sure that AI use in Commission matters is disclosed, that every citation is verified and hyperlinked, that AI-assisted drafts are reviewed by someone with the expertise to catch errors, and that internal AI policies and supervision arrangements reflect the new requirements before 20 October 2026.

A checklist before 20 October 2026

• Identify who in your organisation prepares or contributes to documents lodged in the Commission, and brief them on the three requirements.
• Decide which AI tools may be used for Commission matters, and rule out public tools for anything containing personal or confidential information.
• Build a verification step into the workflow, with a named person responsible for checking facts, citations and quotes against authoritative sources.
• Update templates for responses, submissions and witness statements to include the disclosure and checking statements, and make sure witness statements are drafted from, and confirmed in, the witness’s own words.
• Keep a record of what AI was used for in each document, so that the disclosure can be completed accurately.

If you would like assistance preparing your organisation for the Guidance Note, reviewing your AI use policies, or training your HR and legal teams on the responsible use of AI, please contact a member of our team.