Deloitte faces backlash after admitting to using AI in a government report, resulting in multiple fake citations. The consulting giant is set to refund $440K to the Australian government, sparking debate over ethics, AI use, and accountability in high-stakes consulting.
Thank you for reading this post, don't forget to subscribe!
Global consulting powerhouse Deloitte has agreed to partially refund its $440,000 fee to the Australian government after admitting to using generative artificial intelligence (AI) tools in a government-commissioned report that was later found to contain multiple factual and citation errors.
The report, commissioned in 2024 by the Department of Employment and Workplace Relations (DEWR), was meant to evaluate Australia’s Targeted Compliance Framework. This system issues penalties to welfare recipients who fail to meet their mutual obligation requirements. Instead, the project has become a case study in the risks and responsibilities that come with using AI in high-stakes professional consulting.
What Went Wrong
When Deloitte’s final report was released in July 2025, it identified key flaws in the government’s welfare compliance IT system, describing it as being “driven by punitive assumptions of participant non-compliance.”
But soon after publication, University of Sydney academic Dr Christopher Rudge uncovered glaring inconsistencies. The report included non-existent academic references, fabricated citations, and even a fake quote from a Federal Court judgment.
“These are classic signs of AI hallucinations,”
Rudge said, referring to instances where AI models confidently generate false or misleading information to fill knowledge gaps.
“Instead of replacing fake references with genuine ones, the updated version adds multiple new citations, suggesting the original claims had little evidentiary grounding.”
Deloitte’s Response
Deloitte has since admitted that its researchers used a generative AI large language model (Azure OpenAI GPT-4o) during the early drafting phase of the report. The company insists, however, that human experts “extensively reviewed and refined” the output and that AI had “no impact on the substantive content, findings or recommendations.”
In the updated version of the report, now published on the DEWR website, Deloitte formally discloses the use of AI and corrects more than a dozen references and typographical errors.
A Deloitte spokesperson said the issue has been “resolved directly with the client,” confirming that the refund process is underway.
Government Reaction
The Department of Employment and Workplace Relations emphasized that while errors were corrected, the substance of the review’s findings remained unchanged.
But not everyone was convinced. Labor Senator Deborah O’Neill, a vocal critic of consulting firm ethics, blasted Deloitte’s handling of the situation:
“Deloitte has a human intelligence problem. This would be laughable if it wasn’t so lamentable. A partial refund looks like a partial apology for substandard work.”
She added that public agencies “should be asking exactly who is doing the work they are paying for,” quipping that “procurers would be better off signing up for a ChatGPT subscription.”
Interestingly, Deloitte has recently intensified its AI ambitions. The firm struck a deal with Anthropic to give its nearly 500,000 employees access to the Claude AI chatbot, underscoring how deeply generative AI is being embedded into consulting workflows worldwide.