SYDNEY, Australia (Diya TV) — Deloitte Australia has admitted to using artificial intelligence to help produce a $440,000 report for the federal government — a report that was riddled with errors, fake references, and even a fabricated court quote. The firm has now agreed to issue a partial refund to the Department of Employment and Workplace Relations (DEWR) after a public outcry and an internal review.
The controversial report, first released in July, examined the IT system that automates welfare penalties, such as pauses on unemployment benefits. But it quickly came under scrutiny when University of Sydney academic Dr. Christopher Rudge found multiple false references and mistakes.
Rudge discovered that the report cited nonexistent academic papers, including two attributed to law professor Lisa Burton Crawford and another to Swedish academic Björn Regnell. It also included a fake quote from a Federal Court case, Deanna Amato v. Commonwealth, and even misspelled the name of Justice Jennifer Davies as “Justice Davis.”
Following these revelations, Deloitte conducted an internal review and uploaded a corrected version of the report on Friday, just before a long weekend across much of Australia. The updated version removed the fake references, corrected spelling and typographic errors, and added a significant new disclosure: that generative AI had been used to fill what Deloitte called “traceability and documentation gaps.”
The new version of the report states that Deloitte used “a generative AI large language model (Azure OpenAI GPT-4o) based tool chain licensed by DEWR and hosted on DEWR’s Azure tenancy.” This admission confirmed Rudge’s earlier suspicion that AI-generated “hallucinations” — false or fabricated outputs — had caused many of the errors.
“This is no longer a strong hypothesis,” Rudge said. “Deloitte has now issued a confession, albeit buried in the methodology section. They used generative AI for a core analytical task but failed to disclose this in the first place.”
Rudge warned that the report’s recommendations could not be trusted. “You cannot trust the recommendations when the foundation of the report is built on a flawed, originally undisclosed, and non-expert methodology,” he said.
The case has sparked wider debate about the use of AI in professional consulting and government research. Critics say the incident highlights the risks of relying too heavily on generative AI tools without proper human oversight.
A DEWR spokesperson confirmed that Deloitte had “agreed to repay the final instalment under its contract” but declined to reveal the exact amount. The department maintained that “the substance of the independent review is retained, and there are no changes to the recommendations.”
However, DEWR would not say whether it believed AI was responsible for the report’s inaccuracies. It also declined to confirm whether it plans to continue contracting Deloitte, which has earned nearly $25 million from DEWR since 2021.
A spokesperson for Employment Minister Amanda Rishworth referred all questions to the department.
Internally, Deloitte’s investigation claimed that the errors stemmed from human oversight rather than direct misuse of AI tools. But outside experts remain skeptical.
The incident is particularly damaging for Deloitte because the firm has promoted itself as a global leader in artificial intelligence consulting. Deloitte earns a growing share of its $107 billion annual global revenue from advising companies and governments on AI strategy, ethics, and risk management.
The company also trains executives on how to use AI responsibly and has often emphasized that all AI outputs must be reviewed by humans before publication.
The flawed DEWR report, however, appears to have bypassed that safeguard. The firm’s failure to disclose its AI use — until now — has raised questions about transparency, accountability, and the limits of automation in public-sector consulting.
Despite the controversy, both versions of Deloitte’s report — the original and the revised — identified serious issues in DEWR’s welfare IT system. The report cited poor documentation, undetected system defects, and a punitive compliance model that magnified the effects of system errors on welfare recipients.
These findings align with a previous report from the Commonwealth Ombudsman, which revealed that hundreds of welfare payment suspensions were unlawful.
