Deloitte’s AI fiasco: Australian government refunds contract after chatbot ‘hallucinations’ uncovered

Deloitte has come under fire in Australia after it admitted that sections of a government-commissioned report were written using artificial intelligence — and that the system had fabricated references, quotes, and academic sources.

The incident, which led to Deloitte refunding part of its AU$439,000 (£230,000) consultancy fee, marks one of the first public cases of a government receiving repayment over undisclosed AI use in official work. It has since sparked a national debate about the transparency and credibility of AI-assisted consulting, as per a report by Business Standard.

The controversy began in December 2024, when Australia’s Department of Employment and Workplace Relations (DEWR) hired Deloitte to conduct an “independent assurance review” of its Targeted Compliance Framework, an automated welfare system that penalised jobseekers for missing obligations.

Read More: Bill Gates warns Gen Z: AI alone won’t guarantee you a job

The 237-page report, published in July 2025, initially appeared unremarkable — until Dr Chris Rudge, a researcher at the University of Sydney, discovered that it cited non-existent experts, papers, and even an invented quote attributed to a federal court judge. Some of the supposed studies were linked to the University of Sydney and Sweden’s Lund University, though neither institution had published the works in question.

After media scrutiny, Deloitte confirmed that several references and footnotes were fabricated and admitted that it had used Microsoft’s Azure OpenAI GPT-4o system to assist with drafting parts of the report. The firm subsequently issued a corrected version on 26 September 2025, which now includes a disclosure noting the use of AI tools and removes all falsified citations.

Deloitte has since repaid the final instalment of its consultancy fee following discussions with the department. The DEWR maintained that the report’s core analysis and recommendations remained valid despite the inaccuracies.

Read More: After acquisition, Cognition hits reset at Windsurf with layoffs and buyout offers

The revelation has intensified calls for stricter oversight of AI-generated content in official and corporate work. The timing is also uncomfortable for Deloitte, which only weeks earlier announced a partnership with AI firm Anthropic to give its nearly 500,000 employees access to the Claude chatbot — part of a broader effort by global consulting firms to integrate generative AI into daily workflows.

Read More: Explained: Why Accenture is cutting jobs and what it reveals about workforce transformation