Deloitte was caught using AI in $290,000 report to help the Australian government crack down on welfare after a researcher flagged hallucinations

From the Dossier

Entry 3 of 26. Section A: Firm-specific scandals and accountability failures. Back to the Dossier

Summary

A 237-page “independent assurance” report Deloitte Australia produced for the Department of Employment and Workplace Relations contained fabricated academic citations, a misattributed quote from a federal court judgment, and references to nonexistent studies. University of Sydney researcher Chris Rudge flagged the errors; Deloitte revised the report, disclosed that generative AI (Azure OpenAI) had been used, and refunded part of the government’s payment. Rudge told the AP he “instantaneously knew it was either hallucinated by AI or the world’s best kept secret” upon encountering a fabricated book attribution. The revised report includes a disclosure that Azure OpenAI was used in its creation.

Source description

Fortune’s account is grounded in the department’s own statements and the original Australian Financial Review reporting.

Credibility assessment

This is a textbook case of Rule 3 (Audit the Definitions) failure: the “assurance” report’s definitions of expertise and verification broke down. Under Rule 4, this is an inductive claim about a specific instance, verified through multiple sources including the department’s own statements. The sponsorship context per Rule 6: Deloitte had announced a $3 billion investment in generative AI development, creating a commercial incentive to underplay AI risks.