Fair Work Commission AI Rules: What Employers Must Do Now

Fair Work Commission AI rules took effect on 20 October 2026. Anyone who uses generative AI to prepare a document for a Commission case now has to say so, have a person check it, and declare that the check was done. That includes employers and the HR teams that respond to unfair dismissal and general protections claims. If your business uses ChatGPT, Copilot, Claude or Gemini to draft a response, a submission or a witness statement, your process needs to change this week.

Here is what the guidance note requires, why HR teams are squarely in scope, and a practical checklist for employers and for anyone building AI tools for HR or legal work.

What the new Fair Work Commission AI rules require

The Commission confirmed on 20 October that the requirements in its Guidance note: Use of generative artificial intelligence in Commission cases, released in August, apply from that day in all Commission cases. If you use generative AI to create, write, edit or otherwise prepare a document, three things apply:

1. Disclose it. Tell the Commission that you used generative AI and explain how. Some forms now carry a question about it; for other documents, the Commission suggests a note at the start of the document.
2. Check it yourself. Make sure every fact, piece of evidence, case, piece of legislation and quote is real, accurate and relevant, then state in the document that you have done those checks.
3. Keep witness evidence in the witness’s own words. A witness statement or declaration must be based on the person’s own knowledge, reflect their own words and be true to the best of their knowledge.

Legal practitioners and other professional representatives must also include hyperlinks to any case law referred to in documents prepared using generative AI. According to HRD’s summary of the guidance note, using AI only to check spelling or grammar, or to format a document, falls outside the rules, and the checking must be done by a person: asking the same or a different AI tool to review the document does not count.

The Commission’s GenAI page lists the consequences of not complying: documents can be given less weight or disregarded, a party can be ordered to pay other parties’ costs as well as its own, or a case can be dismissed. It also warns against entering personal or confidential case information into public AI tools.

Why HR teams are in scope

Much of the public discussion has focused on self-represented workers. The rules apply to every party. HRD reports that the guidance note’s definition of a professional representative includes a human resources adviser employed by the party they represent, which means an in-house HR adviser preparing an employer’s response with AI has the same hyperlinking duty as a lawyer.

Brittany Byrne of Citation Legal told HRD that “the rules cut both ways, because unchecked AI-drafted material can be given less weight or disregarded, and non-compliance can lead to costs orders or a case being dismissed.” Tinashe Makamure of Barry Nilsson said even experienced HR teams have been seen quoting a case with the correct name but the wrong citation.

Why the Commission acted

When the guidance note was released in August, the ABC reported on research commissioned by the Commission that found generative AI was partly responsible for a sharp rise in its caseload. The ABC also covered a dismissed Aldi worker who was ordered to pay part of the employer’s legal costs after relying on AI as a “quasi-legal advisor” despite repeated warnings that his case had no substantial prospects of success.

The Commission has not taken an anti-AI line. The ABC reports that it acknowledged AI is helping people with genuine claims who might otherwise have been deterred. Its August announcement came with a research report and practical resources, and its GenAI page now offers set-up prompts that tell an AI tool to rely on the Fair Work Act and official sources, not to invent cases, and to say clearly when the facts do not support a position. The message is that AI is allowed, but a person stays responsible for what is lodged.

A checklist for employers before anything is lodged

1. Find where AI already touches HR documents. Termination letters, investigation reports, chronologies, responses to claims and meeting notes may all pass through Copilot or ChatGPT today. An AI business audit maps those tools and who uses them; here is what an AI audit includes.
2. Record AI use as you go. Add a simple field to your matter file: which tool, what it was asked to do, and who reviewed the output. Writing the disclosure at the end is much harder than capturing it while drafting.
3. Verify every authority by hand. Open each case on AustLII or the Commission’s document search, read the relevant part, and add the hyperlink. Check facts and dates against your own records, not against another AI summary.
4. Leave witness statements to the witness. AI can help organise notes, but the statement must reflect the person’s own knowledge and words. The Commission counts transcribing, translating and editing as uses of generative AI.
5. Keep employee details out of public tools. Use an enterprise or private setup for anything containing personal or confidential case information. Our guide to private LLM options in Australia compares the main approaches.
6. Ground internal HR assistants in your own policies. An assistant that answers from your awards, agreements and policies and cites them makes human checking faster. A RAG knowledge assistant is built that way; see what RAG is.
7. Train the people who draft. Our free AI for HR course covers keeping decisions with people and answering staff from your own policies, and the AI prompts course covers giving AI the right sources and why a person, not a second AI self-check, verifies anything legal before it goes out.

If you build or buy AI for HR or legal work

The guidance note is effectively a product requirement for any tool that drafts workplace or legal documents. Useful features now include:

• Source capture. Every cited case or clause linked to the document it came from, so a person can open it in one click.
• A disclosure record. An automatic log of which parts of a document were drafted or edited with AI, ready to turn into the required statement.
• A human review step. A clear sign-off by a named person before anything leaves the system, rather than an AI “checker” that the Commission would not accept.
• Privacy controls. No personal or confidential case information sent to public models, with access controls and logging.

Our HR AI and people ops automation and legal AI pages show how we approach these builds, and a governed LLM integration layer adds the logging and review queues. For a related example of AI at the front door of legal work, see our look at AI legal intake.

What to watch next

Check the Commission’s GenAI page before each matter, as some Commission forms now include a question about GenAI use. Byrne told HRD she expects disclosure to become routine in workplace disputes, as it is becoming in the courts, and Makamure expects AI-assisted claims to fall only after people face consequences for ignoring the rules. Two other dates matter for HR teams: from 10 December 2026 privacy policies must describe significant decisions made or informed by computer programs (see our guide to automated decision-making and the Privacy Act), and proposals to tighten AI workplace surveillance rules are being debated as the parliamentary AI inquiry hears evidence on AI at work.

This is general information based on Fair Work Commission publications and published reporting, not legal advice. Your obligations depend on your matter and your role in it. Read the Commission’s guidance note and get advice from an employment lawyer about your circumstances.

Featured image: Photo by cottonbro studio on Pexels.