Deloitte to repay Australian government after $440,000 report produced using AI was found with multiple errors

Deloitte has agreed to partially refund the Australian government a $440,000 fee after admitting that it used generative artificial intelligence (AI) to help produce a report that was later found to have multiple errors.

The Department of Employment and Workplace Relations (DEWR) confirmed that Deloitte would repay the final instalment of the contract once the transaction is completed, according to a report by The Guardian.

The department commissioned the consulting firm in 2024 to assess the targeted compliance framework and its supporting IT system, which automatically issues penalties to job seekers who fail to meet their mutual obligation requirements.

According to a report from the news agency AP, the Australian Greens party’s spokesperson, Senator Barbara Pocock, said that Deloitte should refund the entire AU$440,000 ($290,000).

Deloitte “misused AI and used it very inappropriately: misquoted a judge, used references that are non-existent,” Pocock told Australian Broadcasting Corp. “I mean, the kinds of things that a first-year university student would be in deep trouble for.”

Initial report findings and inaccuracies

The report, initially published in July, highlighted major flaws in the system, including “system defects” and a lack of “traceability” between the welfare compliance framework and relevant legislation. It said an IT system was “driven by punitive assumptions of participant non-compliance,” The Guardian reported.

However, the Australian Financial Review (AFR) later exposed multiple inaccuracies within the document, such as non-existent references and fabricated citations, leading to widespread criticism.

University of Sydney academic Christopher Rudge, who first identified the errors, attributed them to AI “hallucinations” where AI models can generate flawed data by trying to fill in gaps or guessing answers. “Instead of just substituting one hallucinated fake reference for a new ‘real’ reference, they’ve substituted the fake hallucinated references, and in the new version, there’s like five, six or seven or eight in their place,” he said.

Deloitte's acknowledgement

The updated version of the report, reuploaded by DEWR on Friday, includes an appendix acknowledging the use of AI tools, specifically mentioning "the use of a generative artificial intelligence (AI) large language model (Azure OpenAI GPT–4o) based tool chain licensed by DEWR and hosted on DEWR’s Azure tenancy.”

Despite the errors, Deloitte maintained that the use of AI did not alter the “substantive content, findings or recommendations” of the report. “The updates made in no way impact or affect the substantive content, findings and recommendations,” the firm said in the amended version.

A spokesperson for Deloitte said, “The matter has been resolved directly with the client”.

However, Labor senator Deborah O’Neill, who was on a Senate inquiry into the integrity of consulting firms, criticised Deloitte by stating, “Deloitte has a human intelligence problem,” she said. “This would be laughable if it wasn’t so lamentable. A partial refund looks like a partial apology for substandard work.”

O’Neill added that public agencies “should be asking exactly who is doing the work they are paying for” and quipped that instead of hiring large consultancies, “procurers would be better off signing up for a ChatGPT subscription," the Guardian report said.