Deloitte Australia has committed to partially refunding the Australian government after its commissioned report was found to contain multiple major errors linked to the use of generative AI.

The 237-page review, commissioned by the Department of Employment and Workplace Relations (DEWR) in December 2024, aimed to assess the IT systems underlying Australia’s welfare compliance program. After publication in July 2025, the report was scrutinised for containing fabricated quotes from a federal court judgment and references to non-existent academic papers.

When challenged by researchers, Deloitte admitted that parts of the document were generated with an AI tool (Azure OpenAI GPT-4o), and corrected the report by revising erroneous citations and footnotes. The firm insists that these corrections do not alter the substantive content, findings, or recommendations of the review.

Still, the case has raised serious concerns about overreliance on AI in high-stakes analysis. Critics—including labor senators and academic observers—have called for stronger review standards, transparency about AI use, and verification of consultancy outputs. 

This episode may become a cautionary benchmark for consulting firms globally: AI can accelerate workflows, but without rigorous oversight, “hallucinations” and factual fabrications risk undermining credibility and trust.