.png)
What Deloitte’s AI Report Controversies Teaches Us About AI Governance
- Vimal Rughani
- Aug 6
- 5 min read
When two government-commissioned reports from the same global consulting network face scrutiny over fabricated citations, false references, and AI-assisted inaccuracies, the conversation must move beyond “AI hallucinations.” It becomes a conversation about governance, accountability, and public trust.
In 2025, Deloitte faced two high-profile controversies involving reports prepared for government entities in Australia and Canada. These incidents are not simply isolated mistakes. Together, they reveal a broader challenge: organizations are adopting generative AI faster than they are building the governance frameworks needed to control it.
For enterprises, regulators, and public institutions, these cases provide a powerful lesson: AI Governance is not an administrative burden — it is the foundation of trustworthy AI.
Case 1: Deloitte Australia and the DEWR report controversy
In October 2025, Deloitte Australia agreed to partially refund the Australian government after a report prepared for the Department of Employment and Workplace Relations (DEWR) was found to contain multiple apparent AI-generated errors.
The 237-page report, valued at A$440,000 (approximately US$290,000), examined IT systems and automated penalties in Australia’s welfare sector.
Confirmed issues in the Australian case
Fabricated academic references
The report cited non-existent academic research papers and books, undermining the credibility of the document.
Why this matters: Government reports must be evidence-based. Fabricated references can mislead policymakers and damage public trust.
Reference: Associated Press reported that the report contained “references to nonexistent academic research papers.”
A fake Federal Court quotation
A quote attributed to a Federal Court judgment was found to be fabricated and did not appear in the actual judgment.
Why this matters: Misquoting a court judgment raises serious concerns about legal integrity and the reliability of AI-assisted research.
Reference: AP News confirmed the report included “a fabricated quote from a federal court judgment.”
Incorrect footnotes and citations
Deloitte reviewed the report and confirmed that some footnotes and references were incorrect.
Why this matters: Incorrect citations violate the principle of traceability, a core requirement of trustworthy AI.
Reference: AP News stated that Deloitte “confirmed some footnotes and references were incorrect.”
Lack of human verification
The errors were not detected during Deloitte’s internal review process. They were identified externally by Chris Rudge, a University of Sydney researcher.
Why this matters: This demonstrates a failure of human oversight, where AI-generated content was not rigorously validated before publication.
Reference: Business Insider reported that the errors were flagged by Chris Rudge after the report had already been published.
Late disclosure of AI use
A disclosure stating that Azure OpenAI GPT-4o had been used in creating the report was added only after the revised version was published.
Why this matters: Transparency about AI usage is essential, especially in high-stakes public-sector documents.
Reference: The Guardian reported that the revised report included a disclosure that Azure OpenAI was used in writing the report.
Case 2: Deloitte Canada and the Newfoundland healthcare report controversy
Shortly after the Australian incident, Deloitte Canada faced scrutiny over a C$1.6 million Health Human Resources Report prepared for the Government of Newfoundland and Labrador.
Public reporting indicated that the report contained false citations and factual inaccuracies likely generated by AI, prompting the provincial government to request a review.
Reported and subsequently acknowledged issues in the Canadian case
False citations
The healthcare report contained false citations that appeared to be generated by artificial intelligence.
Why this matters: False citations undermine the credibility of policy documents that may influence healthcare planning and resource allocation.
Reference: The Independent reported that the report contained “false citations likely generated by A.I.”
Factual inaccuracies
The report was reported to contain numerous factual inaccuracies, raising concerns about the reliability of AI-assisted drafting.
Why this matters: In a healthcare context, inaccurate information can affect policy decisions with real-world consequences for citizens.
Reference: Times of India reported that the report contained factual inaccuracies believed to have been generated by AI tools.
Deloitte’s acknowledgment of AI involvement
Subsequent reporting indicated that Deloitte later admitted the false citations were the result of the use of artificial intelligence.
Why this matters: This reinforces the need for organizations to disclose AI involvement proactively, rather than after problems emerge.
Reference: Canadian Accountant reported that Deloitte later admitted the fake citations were the result of AI use.
A pattern of AI Governance failure
When viewed together, these two cases reveal a pattern of governance weaknesses rather than isolated technical errors.
Lack of source verification
In both cases, AI-generated citations were not independently validated against original sources.
Insufficient human oversight
Errors passed through internal review processes and were identified only after publication or external scrutiny.
Poor traceability
Organizations could not reliably trace every claim back to authentic, verifiable evidence.
Transparency gaps
AI usage was not always disclosed upfront, reducing stakeholder confidence.
Why this matters for AI Governance
These incidents demonstrate that AI Governance is not about restricting innovation. It is about ensuring that innovation is accurate, transparent, accountable, and trustworthy.
A robust AI Governance framework should include:
Mandatory human review of all AI-generated outputs.
Independent source verification for citations, quotations, and references.
Clear disclosure when AI is used in official documents.
Audit trails that record prompts, outputs, edits, and approvals.
Risk classification for high-stakes AI use cases.
Defined accountability for the final accuracy of AI-assisted deliverables.
The broader implication for enterprises
The Deloitte controversies are particularly significant because they involved government-commissioned reports. When AI-generated inaccuracies enter public policy documents, the consequences can extend beyond reputational damage to affect governance, regulation, healthcare planning, and public trust.
For enterprises adopting AI in ERP, finance, healthcare, legal, and consulting functions, the lesson is clear:
AI can accelerate research, drafting, and analysis, but it cannot replace human accountability.
Without strong AI Governance, organizations risk publishing false information, misleading stakeholders, and damaging the trust they have spent years building.
Conclusion
Deloitte’s AI report controversies in Australia and Canada should be viewed as a global wake-up call for enterprise AI Governance.
The technology itself is not the enemy. The real risk lies in deploying AI without the governance mechanisms needed to verify its outputs, disclose its use, and assign accountability for its mistakes.
As AI becomes deeply embedded in enterprise systems and public-sector decision-making, organizations must ensure that every AI-generated claim is transparent, traceable, and accountable.
The future of AI will not be determined solely by how powerful the technology becomes. It will be determined by how responsibly we govern it.
References
Associated Press. Deloitte to partially refund Australian government for report with apparent AI-generated errors. 7 October 2025. https://apnews.com/article/ab54858680ffc4ae6555b31c8fb987f3
Business Insider. Deloitte is giving the Australian government a partial refund after it used AI to deliver a report with errors. 6 October 2025. https://www.businessinsider.com/deloitte-australia-issues-refund-ai-assurance-project-2025-10
The Guardian. Deloitte to pay money back to Albanese government after using AI in $440,000 report. 6 October 2025. https://www.theguardian.com/australia-news/2025/oct/06/deloitte-to-pay-money-back-to-albanese-government-after-using-ai-in-440000-report
The Independent. Provincial government responds to Deloitte report found to contain false citations likely generated by A.I. 22 November 2025. https://theindependent.ca/news/lji/provincial-government-responds-to-deloitte-report-found-to-contain-false-citations-likely-generated-by-a-i/
The Independent. Major N.L. healthcare report contains errors likely generated by A.I. 22 November 2025. https://theindependent.ca/news/lji/major-n-l-healthcare-report-contains-errors-likely-generated-by-a-i/
Canadian Accountant. Newfoundland and Labrador: Province updated public contracting templates after Deloitte AI citation scandal. 18 March 2026. https://www.canadian-accountant.com/content/profession/cpanl-deloitte-ai-scandal
Times of India. Deloitte makes AI mistake again, this time in report for Canadian government; here’s what went wrong. 7 November 2025. https://timesofindia.indiatimes.com/technology/tech-news/deloitte-makes-ai-mistake-again-this-time-in-report-for-canadian-government-heres-what-went-wrong/articleshow/125585971.cms
Comments