Deloitte To Repay Australian Government After AI Errors Found In Official Report

Global consulting giant Deloitte has agreed to refund a part of its $440,000 (($290,000) fee to the Australian government after admitting to using generative AI tools in a report assessing the government's "Future Made in Australia" initiative. The department commissioned the company in 2024 to review the targeted compliance framework and its associated IT system, which automatically issues penalties to job seekers who do not meet mutual obligation requirements, Guardian reported.

The final report, released in July, was found to contain several significant errors -- including academic citations referencing individuals who do not exist and a fabricated quote from a Federal Court judgment, according to a report by the Australian Financial Review.

On Friday, the Department of Employment and Workplace Relations published an updated version of the report on its website. The revised copy removed over a dozen fictitious references and footnotes, updated the reference list, and corrected numerous typographical error. 

Australian welfare academic Dr Christopher Rudge, who first spotted the errors, said the report contained AI "hallucinations"--instances where AI models generate false or misleading information by filling in gaps, misinterpreting data, or guessing answers.

"Rather than simply replacing a single fake reference with a real one, they've removed the hallucinated citations and, in the updated version, added five, six, even seven or eight new ones in their place. So what that suggests is that the original claim made in the body of the report wasn't based on any one particular evidentiary source,” he said.

What did Deloitte say?

Deloitte acknowledged the use of AI but stated it was only a tool used during the early drafting stages and that the final product was extensively reviewed and refined by human experts. The company insisted that the use of AI had no impact on the "substantive content, findings or recommendations" of the report. While Deloitte acknowledged the use of generative AI tools, it did not attribute the errors in the original report directly to artificial intelligence. 

In the new version, Deloitte also formally disclosed that its research methodology included the use of a generative artificial intelligence tool, specifically, a large language model (Azure OpenAI GPT-4o). 

A spokesperson for Deloitte confirmed that "the matter has been resolved directly with the client." The Department has since confirmed that the refund process is underway and that future consultancy contracts may include stricter guidelines on the use of AI-generated content. 

Ethical Implications of Using AI

The incident has sparked a wider debate on the ethical and financial implications of using artificial intelligence in high-value consultancy work. With firms increasingly turning to AI for speed and efficiency, questions are being raised about how much of that work is human-led and whether clients are getting fair value. 

Interestingly, Deloitte recently signed a deal with Anthropic to give its nearly 500,000 global employees access to the Claude chatbot, highlighting the growing reliance on AI in professional services. 

This case marks one of the first major instances in Australia where a private firm has faced consequences for undisclosed AI usage in government-funded work.