Should AI Write Work Emails? What Australian Tribunals Ruled in 2026

An Australian knowledge worker writes a workplace email in a modern office. An open laptop, stacked textbooks, coffee cup, and small potted fern adorn their wooden desk. The real wall is embossed with the scales of justice and a courthouse silhouette.

In 2026, the Fair Work Commission started ruling on a problem most workplaces haven't caught up to yet: what happens when someone lets AI write the email, the complaint or the legal submission, and nobody checks it before it goes out. Three cases this year show the pattern, and the Commission has now written a guidance note to deal with it directly, effective from 20 October 2026.

Three Fair Work Commission cases about AI-written communication

A bullying complaint buried by its own AI-generated length

A Queensland worker's bullying complaint against her employer grew into pages of AI-drafted correspondence. Personnel Today reported the complaint was effectively buried by its own volume, with the tone and length driving the reader to stop before reaching the genuine grievance underneath. Commissioner Sarah McKinnon found the correspondence "lengthy, wide-ranging, repetitive, and couched in accusatory language," and ordered that all future communication between the two parties exclude AI tools entirely: "There is to be no use of artificial intelligence tools in the preparation of correspondence" (ABC News, 31 August 2026).

A software developer's AI-drafted emails helped sink his unfair dismissal claim

Marcus Wibmer, a senior Java developer at Fujifilm Data Management Solutions in Sydney for over a decade, was dismissed in August 2025 after a workplace dispute escalated into a run of AI-drafted emails to multiple managers. The Australian Computer Society reported the Fair Work Commission described his messages as "dense, repetitive and often rambling," pulling in complaints about dress code, sexual harassment and whistleblower protections that had little to do with the original disagreement. Deputy President Tony Slevin found AI had given Wibmer false confidence in what he was sending: the tool "appears to have given [him] a false sense of security" (HR Daily, 20 March 2026). The Commission upheld his dismissal.

An Uber driver's AI-written submissions dismissed as incoherent

In Zain Anwar v Rasier Pacific Pty Ltd, an Uber driver deactivated over four incidents of misconduct involving customers used generative AI to draft his applications, emails, submissions and witness statements to the Commission without reviewing them first. The Full Bench found his submissions "by and large, incoherent" and named him part of "an increasing trend of applicants misusing generative artificial intelligence to run their cases." Every application was dismissed. The bench's reasoning was plain: "Generative AI is merely a tool. It is not capable of making the type of decisions required to effectively conduct litigation" (Fair Work Commission, 19 June 2026).

Why Fair Work case numbers are climbing

Overall Fair Work Commission lodgments rose 44 per cent between the 2022-23 and 2024-25 financial years, a jump Commission President Justice Adam Hatcher has linked to the growing use of generative AI to prepare claims (SmartCompany, 26 August 2026). Hatcher has separately put the Commission's annual caseload at around 55,000 matters in the 2025-26 financial year, up from a "normal" 30,000 as recently as 2023 (Kingston Reid, 18 May 2026).

Griffith University senior lecturer Dr Mai Nguyen says she now suspects about 80 per cent of the emails landing in her own inbox are AI-generated, and that the longer ones are usually the giveaway (ABC News, 31 August 2026). Australia has no dedicated AI Act, so for now it's tribunals like the Fair Work Commission working out the rules case by case, according to University of Melbourne AI governance researcher Dr Guzyal Hill (ABC News).

The Fair Work Commission's new rules on AI, from 20 October 2026

From 20 October 2026, a new Fair Work Commission guidance note requires anyone using generative AI to prepare a case document, an application, submission or witness statement, to disclose that use and how it was used (Fair Work Commission; SmartCompany). They also have to check that every fact, quote and legal reference in the document is correct, and witnesses must independently confirm a statement reflects their own knowledge, not the AI's draft. The guidance separately warns against feeding personal or confidential case details into public AI tools such as ChatGPT, Claude, Copilot or Gemini. Undisclosed AI use can see a document given less weight or thrown out, and the Commission can order the non-disclosing party to pay the other side's costs.

What these cases have in common

In each case, AI drafted fluent text that nobody then checked against how the recipient would read it. Wibmer's emails expanded a single complaint into a string of unrelated allegations. Anwar's submissions repeated the same claims until the Commission could no longer follow them. The caretaker's complaint had legitimate concerns inside it, and lost them under the volume AI produced around them. QUT digital communication professor Daniel Angus has flagged the same failure mode in adversarial workplace settings: AI's fluency lets a sender bury the other side in length rather than resolve anything (ABC News).

Where general AI tools fall short, and what Unanimas is building instead

Most AI tools on the market are built to generate words fast: draft the email, clear the inbox, tidy the grammar. None of that checks whether the tone and intent behind the message will land the way the sender means it, which is the gap showing up across these rulings. The words read as normal to the person who sent them, and as something else entirely to the tribunal reading them back.

Unanimas is building communication intelligence for the AI era. Where today's AI helps people generate words, clear inboxes or check grammar, Unanimas is working on helping people understand and be understood. Harmony, its first product, brings that to workplace email and chat: it works inside the tools a team already uses, and it translates a sender's intent into a message the recipient can hear, rather than generating new claims or arguments on the sender's behalf. Sensitive material is sanitised locally before anything leaves the environment, a promise Unanimas calls Data Serenity™ (unanimas.com).

Harmony doesn't draft a Fair Work Commission submission for anyone, and it isn't built to. What it's built for is the moment before any of that: the email or message someone is about to send in the middle of a workplace dispute, before it turns into fifteen pages a Commissioner has to read.

FAQs: AI and Australian workplace communication in 2026

Can I get fired for using AI to write work emails in Australia?

Not on its own, but 2026 Fair Work Commission rulings show AI-drafted correspondence can count against an employee when it's excessive, inaccurate or wrongly toned. In Marcus Wibmer's case, the Commission upheld his dismissal partly because his AI-drafted emails were inappropriate for a workplace (HR Daily; ACS).

What are the Fair Work Commission's new rules on generative AI?

From 20 October 2026, anyone using generative AI to prepare a Fair Work Commission document must disclose that use and verify the content is accurate. Witness statements need independent confirmation that they reflect the witness's own knowledge, not the AI's wording (Fair Work Commission).

Can AI make up case law in a Fair Work Commission claim?

Yes. The Commission has found AI tools can generate fabricated legal citations that read as genuine, which is why every AI-assisted filing must now be checked before it's submitted (SmartCompany; Kingston Reid).

Does the Fair Work Commission ban AI in workplace disputes?

Not outright, but in one case a Commissioner ordered AI tools excluded entirely from future correspondence between two parties after AI-generated complaint content became unmanageable (ABC News; Personnel Today).

Is Harmony the same as using LLMs or AI writers for work emails?

No. General-purpose AI tools generate new text from a prompt. Harmony works from what someone has already written, and adjusts the tone so it matches what they mean, inside the email and chat tools a team already uses (unanimas.com).

{{ include_custom_fonts({"DM Sans":["Regular","Regular Italic"]}) }}