Deloitte Australia Issues Partial Refund Over AI-Generated Inaccuracies in Government Report
MELBOURNE, Australia – Financial services giant Deloitte Australia has agreed to issue a partial refund of the AU$440,000 (approximately $290,000 USD) paid by the Australian government. This decision follows the discovery of numerous apparent AI-generated inaccuracies in a 237-page report, which reviewed the Department of Employment and Workplace Relations’ IT systems concerning automated penalties in the welfare system. The report was found to contain fabricated court judgments and references to nonexistent academic research.
Discovery of “Fabricated References”
The significant errors within the report came to light after Chris Rudge, a researcher specializing in health and welfare law at Sydney University, alerted media outlets. Rudge described the initial version of the report, published in July, as “full of fabricated references.” He reportedly identified up to 20 inaccuracies, with the first glaring error being a wrongly attributed book to Sydney University professor Lisa Burton Crawford—a work Rudge immediately recognized as preposterous and outside her known field of expertise.
A more serious concern for Rudge was the misquotation of a federal court judge, a direct misrepresentation of legal standing in a document intended to audit the government’s legal compliance. “They’ve totally misquoted a court case then made up a quotation from a judge,” Rudge stated, emphasizing the critical nature of misstating the law to the Australian government.
Deloitte’s Response and Revised Publication
The Department of Employment and Workplace Relations confirmed on Tuesday that Deloitte had reviewed the report and “confirmed some footnotes and references were incorrect.” As a result, Deloitte agreed to repay the “final installment” under its contract. While the exact amount of the refund has not yet been publicly disclosed, it will be made known once reimbursed.
Following the revelations, a revised version of the report was published on Friday. This updated document notably included a disclosure that a generative AI language system, Azure OpenAI, had been used in its creation. Removed from the revised report were the fabricated quotes attributed to a federal court judge and references to nonexistent reports by law and software engineering experts.
When pressed for comment on whether the errors were specifically generated by AI, Deloitte merely stated that the “matter has been resolved directly with the client,” declining to confirm or deny AI’s direct role. The phenomenon of generative AI systems producing false or misleading information is widely known as “hallucination.”
Calls for Full Refund Amid AI Concerns
Despite the identified inaccuracies, the department stated that the “substance” of the report and its recommendations remained unchanged. However, the incident has drawn sharp criticism from Australian lawmakers.
Senator Barbara Pocock, the Australian Greens party’s spokesperson on the public sector, has called for Deloitte to refund the entire AU$440,000. Pocock condemned Deloitte’s actions, stating, “Deloitte misused AI and used it very inappropriately: misquoted a judge, used references that are non-existent… I mean, the kinds of things that a first-year university student would be in deep trouble for.”
The incident underscores growing concerns about the responsible integration of artificial intelligence into professional services and the necessity for rigorous human oversight to ensure accuracy and maintain public trust.
