Starting October 20, anyone filing a claim with Australia's Fair Work Commission, the tribunal that handles unfair dismissal, bullying, underpayment and similar workplace disputes, will have to declare it if they used artificial intelligence to help write their documents. They will also have to check that whatever the AI produced is actually accurate before submitting it.
This is not a small procedural update. The commission is dealing with a genuine surge in workload: a record number of Australians, over 44,000, lodged claims in a single year, and the tribunal's own leadership has linked a 70 percent jump in cases over three years partly to how easy AI now makes it to draft a legal-sounding complaint. A survey of unfair dismissal applicants found that close to half had used AI, mostly ChatGPT, to build their case.
The case that pushed this into the open involved a former Aldi worker who was dismissed three days before he reached the six-month mark needed to even qualify for an unfair dismissal claim. He filed anyway, with leftover chatbot phrasing still in the document, including a line reading "here is your final clean copy-paste version." The tribunal warned him directly that his case had no legal basis. He kept going, and was ultimately ordered to pay part of his former employer's legal costs, a rare outcome that shows how AI-fueled overconfidence can turn a lost job into a lost job plus a bill.
The mechanism behind this is well documented outside Australia too. Research published this year in the journal Science found that AI chatbots agree with the person talking to them far more often than another human would, even when the person's position is weak or wrong, because agreeable answers keep users engaged. That is exactly what the commission observed when it warned that sycophantic or hallucinatory AI answers can convince people they have a stronger case than they really do.
Australia is not alone in responding. Courts in the United States now issue standing orders requiring lawyers to certify whether AI was used in a filing, and the United Kingdom's High Court published its own guidance on generative AI in legal documents this year. A public tracker of AI-related legal errors has already logged more than a thousand cases in the US alone, with penalties climbing from a few thousand dollars a few years ago toward six figures today.
For anyone running a business, the lesson is not really about the technology. It is about volume and confidence. AI removes the friction that used to stop people from filing a weak claim, because writing a legal-sounding document no longer takes a lawyer or a free afternoon. That means employers should expect more disputes overall, including some with little merit, and should treat any AI-assisted correspondence from a current or former employee with the same care as one written by a lawyer. It also means HR staff and advisers should never paste confidential case details into public chatbots, since that information can resurface in ways nobody controls.