Consulting firm Deloitte will refund part of a A$440,000 (£216,000) fee to the Australian government after a report it produced using artificial intelligence was found to contain serious factual errors.
The report, commissioned by the Department of Employment and Workplace Relations (DEWR), aimed to assess the “Future Made in Australia” compliance framework and its IT systems.
Published in July 2025, it was later found to include fabricated academic citations, false references, and even a quote wrongly attributed to a Federal Court judgment.
Deloitte has since admitted that it used Microsoft’s Azure OpenAI GPT-4o model to help draft sections of the report. The firm said that human reviewers refined the AI’s output and insisted that the report’s main findings and recommendations remained accurate.
Errors and Refund
After the issues were raised, the government ordered Deloitte to revise the report. More than a dozen fake references were removed or replaced, and several typographical errors were corrected.
The department confirmed that Deloitte would refund the remaining part of its consultancy fee once the process is completed.
Sydney-based academic Christopher Rudge, who first spotted the errors, said the false citations were examples of “AI hallucinations”, when a generative model produces convincing but entirely made-up information.
Senator Deborah O’Neill, of the Australian Labor Party, criticised Deloitte, suggesting the firm had “a human intelligence problem”. She added that the incident raised questions about the government’s reliance on large consulting firms.
A Global Pattern of Scrutiny
While this is one of the first cases where a consultancy has been forced to repay fees due to AI errors, Deloitte and other major firms have faced increasing regulatory pressure worldwide.
In India, the National Financial Reporting Authority fined Deloitte Haskins & Sells Rs 2 crore in 2024 for audit lapses at Zee Entertainment. The company’s affiliates in China, Colombia and Canada have also been penalised in recent years for separate auditing failures.
In 2022, US regulators fined Deloitte’s Chinese arm US$20 million for allowing clients to audit their own work, a breach of professional ethics. In Colombia, the firm’s affiliate was fined US$900,000 for quality control violations, and in Canada, Deloitte paid C$1.59 million after admitting to deliberately backdating audit paperwork.
Although none of these cases involved AI, experts say they highlight growing concerns about transparency and accountability in the professional services industry.
Lessons for the Future
Experts warn that generative AI tools can produce realistic but false information if not carefully monitored. They argue that firms under pressure to meet tight deadlines may be tempted to rely too heavily on AI to speed up work.
Analysts say the Deloitte case is a reminder that human oversight and fact-checking remain essential, especially when dealing with high-stakes government or financial reports.
The Australian government has indicated that future consultancy contracts are likely to include stricter rules on when and how AI can be used.
As organisations worldwide embrace AI technology, the Deloitte incident stands as a cautionary tale, showing that while AI can boost productivity, it can also damage trust when accuracy and accountability are overlooked.
