Deloitte used AI to generate false academic citations in a $1.6 million Canadian government healthcare report, creating fictional research papers and misattributing work to real researchers who had never collaborated.
In May 2025, the Canadian province of Newfoundland and Labrador published a 526-page Health Human Resources Plan commissioned from consulting firm Deloitte for nearly $1.6 million. An investigation by The Independent found that the report contained at least four false citations that appear to be AI-generated. These included fictional academic papers with made-up titles like 'The cost-effectiveness of a rural retention program for registered nurses in Canada' and 'The cost-effectiveness of local recruitment and retention strategies for health workers in Canada.' Real researchers named in these citations confirmed they had never written such papers or worked with all the co-authors listed. One citation referenced a non-existent article in the Canadian Journal of Respiratory Therapy. Deloitte admitted that 'AI was not used to write the report; it was selectively used to support a small number of research citations.' The report was used to advise the government on healthcare workforce planning, including virtual care, retention incentives, and COVID-19 impacts on healthcare workers during critical staffing shortages. This incident follows a similar case where Deloitte's Australian office used AI in a $290,000 government report that also contained fabricated citations and had to provide a partial refund.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that inadvertently generate or spread incorrect or deceptive information, which can lead to inaccurate beliefs in users and undermine their autonomy. Humans that make decisions based on false beliefs can experience physical, emotional or material harms
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed