When a high-profile government report turned out to be riddled with errors and imaginary citations, Australia’s Department of Employment and Workplace Relations found itself in an unexpected scandal — not because of politics, but because of artificial intelligence.
A costly contract gone wrong
It began as a serious undertaking: the Australian government commissioned consulting giant Deloitte to produce a 273-page report on how automation and artificial intelligence could improve the administration of welfare penalties. The contract, valued at 440,000 Australian dollars (about €250,000), was meant to guide policy on one of the country’s most sensitive social issues.
But the polished report delivered to the Department of Employment and Workplace Relations quickly came under scrutiny. Academics reviewing the document noticed something strange — citations that didn’t exist. One health law professor from the University of Sydney pointed out several references to scholarly research that had never been published, including supposed work by real academics who had, in fact, never written such papers.
In other words, large portions of the report had been generated by an AI tool — apparently without any verification.
The AI “hallucination” problem
According to multiple reports, the document contained “hallucinations,” a term used in AI research to describe fabricated or misleading content produced by language models. These errors often appear convincing at first glance but collapse under factual scrutiny.
Deloitte later admitted that parts of the report were produced using a “generative language model toolchain (Azure OpenAI GPT-4o)”, confirming that AI had been used to draft and reference sections of the study. The company claimed that the errors were limited to a “small number” of footnotes and references and that corrections had since been made.
Still, the revelation sparked public outrage — not least because of the hefty price tag attached to the flawed document.
Deloitte’s refund and public backlash
Following mounting criticism, Deloitte agreed to refund a portion of the contract, though it declined to specify the exact amount. The consulting firm described it as a “good faith gesture” and maintained that the core findings and recommendations of the report remained “accurate and valuable.”
Government officials were more cautious. While the Department said the “independent audit’s substance remains intact,” it acknowledged that the use of generative AI “was not made transparent at the time of publication.”
Critics, however, weren’t so forgiving. As one observer put it, “When taxpayers fund a six-figure research project, they expect real experts — not a chatbot — to be doing the analysis.”
A warning for the AI age
This incident is more than a bureaucratic embarrassment; it’s a case study in the risks of over-relying on artificial intelligence without proper oversight. Experts like Dr. Toby Walsh, a leading AI researcher at the University of New South Wales, have long warned that “AI can generate convincing nonsense,” urging governments and businesses to use such tools responsibly and transparently.
In an era when public institutions increasingly lean on automation and machine learning, the Australian “AI report fiasco” serves as a wake-up call: technology may be powerful, but without human verification, even the most sophisticated systems can lead to costly — and embarrassing — mistakes.
At least this time, Australia got a partial refund.
