Professional services and consulting firm Deloitte has said it will issue an unspecified partial refund to the Australian federal government after it was found to have used AI to prepare a report for which it charged $440,000 (US$291,000).
The Department of Employment and Workplace Relations (DEWR) commissioned the report in 2024 to assess its targeted compliance framework and its supporting IT system, which automatically issues penalties to job seekers who fail to meet their mutual obligation requirements.
The first version was published in July this year. About a month later, Christopher Rudge from the University of Sydney highlighted multiple errors in the document and speculated they might be the result of an AI “hallucination,” which occurs when the technology responds to user queries by inventing references and quotes.
Rudge pointed out that Deloitte’s analysis included three non-existent academic references and an invented quote supposedly from a federal court judgment.
The company refused to comment, but a new version was quietly uploaded to the Department of Employment and Workplace Relations (DEWR) website on Oct. 3, containing more than a dozen deletions of non-existent references and footnotes, a new reference list, and other corrections.
Despite being partially based on false data, Deloitte’s recommendations were unchanged from the first draft.
The new version admits that the firm used “a generative AI large language model (Azure OpenAI GPT-4o) ... licensed by DEWR and hosted on DEWR’s [servers]” for what it called “traceability and documentation gaps.”
The minister responsible, Amanda Rishworth, has refused to comment and referred questions to DEWR, which confirmed that the consultants had “agreed to repay the final instalment under its contract.”
Meanwhile, a spokesperson for Deloitte said, “The matter has been resolved directly with the client.”
But New South Wales Labor Senator Deborah O'Neill, who was a leading figure in the parliamentary inquiry into the PwC tax leaks scandal, says that’s insufficient.
“A partial refund looks like a partial apology for substandard work,” she told The Australian Financial Review. “Deloitte has a human intelligence problem. This would be laughable if it wasn’t so lamentable ... too often, as our parliamentary inquiries have shown, these consulting firms win contracts by promising their expertise, and then when the deal is signed, they give you whatever [staff] costs them the least.
Auditors Failing to Audit Their Own Use of AI: UK Regulator
In June, the UK’s accountancy regulator, the Financial Reporting Council (FRC), raised concerns (pdf) over the failure to measure how artificial intelligence and automated tools are affecting audit quality, despite widespread use across the profession and amongst consultancies like McKinsey.It said that while these firms track usage—typically for licensing or rollout purposes—none, bar one unnamed firm, had defined key performance indicators (KPIs) to monitor the tools’ contribution to audit quality.
“AI tools are now moving beyond experimentation to becoming a reality in certain audit scenarios,” said Mark Babington, executive director of regulatory standards at the FRC.
He warned that the benefits will only be realised “if tools produce consistently reliable outputs and are used routinely in the intended manner.”
Along with the other members of the “Big Four”—PricewaterhouseCoopers (PwC), Ernst & Young (EY), and KPMG—Deloitte has been heavily investing in AI.
On the same day it was revealed the company would issue the government a refund over AI-produced slop, it signed an enterprise agreement with Anthropic, developers of a family of large language models (LLMs) named Claude.
Deloitte also plans to create different AI agent “personas” to interact with customers. They will represent the various departments within the company, including accountants and software developers.
“Deloitte is making this significant investment in Anthropic’s AI platform because our approach to responsible AI is very aligned, and together we can reshape how enterprises operate over the next decade. Claude continues to be a leading choice for many clients and our own AI transformation,” Ranjit Bawa, global technology and ecosystems and alliances leader at Deloitte, said in the blog post.
