MELBOURNE, AUSTRALIA — Deloitte has agreed to refund the Australian government AUD 440,000, following errors in its AI-assisted report.
The report, commissioned by the Department of Employment and Workplace Relations (DEWR), was intended to be an assurance review of Australia’s Targeted Compliance Framework. It was initially published in July on the department’s website, but was flagged due to numerous issues:
- False academic references
- A fabricated Federal Court judgment quote
Deloitte reviewed the 237-page report and issued a revised version that included a disclosure that part of its methodology used generative AI, specifically Azure OpenAI GPT-4. Federal court quotes and nonexistent citations were also removed. However, the firm asserts that the substantial content, findings, and recommendations remain unchanged after the updates.
Fueling the Concerns Over AI Hallucinations
I instantaneously knew it was either hallucinated by AI or the world’s best-kept secret because I’d never heard of the book, and it sounded preposterous.
They’ve totally misquoted a court case, then made up a quotation from a judge, and I thought, well, hang on: that’s actually a bit bigger than academics’ egos. That’s about misstating the law to the Australian government in a report that they rely on. So, I thought it was important to stand up for diligence.
— Chris Rudge, Sydney University Researcher
The case was an evident warning shot against firms’ reliance on generative AI tools, highlighting the risk of AI hallucinations—the tendency for AI to produce plausible but fabricated information. This prompted a call for stronger ethical AI auditing protocols and for firms to find the balance between innovation and oversight.
