Western Australia demands refund from Deloitte over AI errors in university merger report

Errors in Deloitte’s university merger report

Western Australia’s government commissioned Deloitte to produce a cost-benefit analysis for merging the University of Western Australia, Curtin University, and Murdoch University. The report, which cost A$166,000, was later found to contain AI-generated references to non-existent sources and incorrect financial totals.

Critics first raised concerns in October after a Reddit post highlighted discrepancies. Guardian Australia subsequently identified errors in the report’s appendix, including a misreported total where two costs for a merger between Curtin University and the University of Western Australia were listed as A$399m instead of A$659m.

Deloitte acknowledged in the report that it used generative AI to gather research, summarise information, and proofread. The government stated that while the errors were disappointing, they do not affect the report’s overall modelling or findings.

Government response and refund demand

The government has written to Deloitte to expedite a refund and asked the firm to correct and reissue the report. A spokesperson confirmed that Deloitte will address the errors but did not comment further on the refund request.

This is not the first time Deloitte has faced scrutiny over AI-generated errors. In 2025, the firm partially refunded Australia’s federal government for mistakes in a A$440,000 report. Rival firms KPMG and PwC have also encountered similar issues with AI-written documents this year.

Broader context for consulting and AI

The incident follows a pattern of consulting firms grappling with the risks of generative AI in high-stakes reports. Earlier this year, LazyFounder reported on PwC’s alliance with Cohere to deliver sovereign AI solutions, a partnership that underscores the growing integration of AI in consulting services. However, this case highlights the potential pitfalls of relying on AI for critical research and analysis.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

For founders and operators, this story underscores the risks of integrating generative AI into high-stakes consulting or research work. While AI can streamline processes, errors like those in Deloitte’s report demonstrate that human oversight remains essential—especially when outputs inform government decisions or financial modelling. The demand for a refund also signals that clients, particularly public-sector ones, are increasingly unwilling to accept AI-generated mistakes in costly contracts. Consulting firms and startups building AI-driven tools must prioritise accuracy and transparency to avoid reputational damage and financial penalties.

Key takeaways

• AI-generated errors in consulting reports can lead to financial and reputational consequences, as seen with Deloitte’s refund demand.
• Governments and clients are becoming less tolerant of mistakes in AI-assisted work, particularly when contracts involve significant public funds.
• Human review remains critical for validating AI outputs, especially in high-stakes reports or modelling.
• This incident adds to a growing list of AI-related missteps by consulting firms, reinforcing the need for robust quality control measures.

Related on Lazyfounder

• PwC
• PwC and Cohere ally to sell sovereign AI, starting in Canada
• Reddit to shut down RSS feeds and public API access, citing AI scraping concerns