The Department of Employment and Workplace Relations (DEWR) has confirmed that Deloitte Australia will refund part of a A$440,000 government contract after a department-commissioned report was found to contain AI-generated errors — including fabricated citations and a non-existent court judgment.
The “Targeted Compliance Framework Assurance Review,” published in July 2025, was meant to assess welfare compliance processes. But soon after its release, University of Sydney researcher Dr Christopher Rudge discovered that sections of the report cited sources that did not exist. Following an internal review, Deloitte admitted generative AI tools were used during drafting, producing false references that went unchecked before publication.
DEWR later uploaded a corrected version of the report in October, removing the fabricated material. The department said the “substance of the review remains intact” and that its recommendations were unaffected. Deloitte agreed to forgo the final instalment of its contract and partially refund the government, while defending the report’s overall conclusions.
The incident has sparked widespread debate about the unchecked use of AI in government consulting work. Senator Deborah O’Neill described it as “a wake-up call on transparency and accountability” for firms providing advice to the public sector. Parliamentary committees are now examining whether government contracts should require mandatory disclosure when AI tools are used in preparing reports.
Deloitte, one of Australia’s “big four” consultancies, said it has strengthened internal review processes and reaffirmed that human experts retain responsibility for final deliverables. DEWR has not announced any new work with Deloitte since the refund but said it would continue reviewing procurement guidelines to ensure “robust assurance over the use of emerging technologies.”
The scandal has become the clearest example yet of the reputational risks posed by generative AI in high-stakes policy reporting — and a lesson in the need for rigorous human oversight in the age of automated drafting.
When AI makes things up: Deloitte’s costly lesson in artificial intelligence hallucinations
