Deloitte caught using hallucinating AI in report for Australian Government

Deloitte has been forced to return part of a AU$439,000 (around €260,000) fee to Australia’s Department of Employment and Workplace Relations (DEWR) after admitting that a report it produced contained multiple factual and citation errors – some of which were generated by artificial intelligence.

The “independent assurance review,” commissioned in December 2023 to assess problems within the Government’s welfare compliance framework, was published earlier this year before being quietly replaced with a corrected version last week.

The revised report acknowledged that a “generative artificial intelligence (AI) large language model (Azure OpenAI GPT-4o) based tool chain” had been used in the preparation of the document – a detail not disclosed in the original version.

The Australian Financial Review first uncovered that the report contained more than a dozen fake references, including academic papers and books that did not exist, and even a fabricated quote from a Federal Court ruling. The inaccuracies were first spotted by academic Chris Rudge, who cross-checked the citations.

Following the revelation, Deloitte updated the report, correcting the footnotes and bibliography while maintaining that “the updates made in no way impact or affect the substantive content, findings and recommendations.”

However, the firm has agreed to refund the final instalment of its contract as part of an agreement with the Government department. A DEWR spokesperson confirmed the arrangement, adding that the amended version “did not change the substance or overall recommendations” of the review.

The episode highlights growing concerns over the use of generative AI within the consulting and audit sectors. While AI tools can dramatically speed up research and drafting processes, they are also prone to “hallucinations” – the confident invention of false or unverifiable information.

The Financial Times noted that the UK accountancy regulator had already warned in June that major audit firms were failing to adequately track how automation and AI tools affected the quality of their work, even as their use expanded rapidly in risk assessment and evidence-gathering.

Deloitte’s admission underscores how these technologies – when deployed without sufficient human oversight – can expose even the largest firms to reputational and financial risks.

Industry observers say this case reflects a broader shift in how large consultancies operate.

Commenting on LinkedIn, Oriane Cohen, Founder of the Grey Zone and OC Strategic Advisory, wrote that Deloitte’s report “is not just a bug – it’s the business model,” arguing that big firms increasingly rely on junior staff and AI tools to deliver reports at premium rates.

“These firms sell access and brand reputation,” Ms Cohen said. “But the real intelligence work happens elsewhere – in smaller, specialised firms where leadership works directly on your case and quality control doesn’t rely on hoping AI didn’t hallucinate.”