A health care report prepared by Deloitte Canada for a Canadian provincial government is facing serious criticism after errors were discovered in its research. The report, which cost the province nearly $1.6 million, was published earlier this year and aimed to guide government decisions on health care.
The 526-page study covered important topics such as virtual care, staff retention, and the impact of COVID-19 on doctors and nurses. These topics are especially significant because the province is currently struggling with a shortage of medical staff.
Investigators found that the report contained false citations. Some research papers listed in the report do not exist, while some real researchers were credited for papers they did not write. In other cases, authors were listed together even though they had never worked on a paper jointly.
The errors raise concerns about how the report was prepared and whether it can be fully trusted to guide government decisions.
Possible AI involvement
Deloitte Canada maintained that AI tools were not used to write the report in full, but acknowledged that AI was employed to assist with a limited number of research citations. Analysts who examined the report noted that the nature of the errors—such as references to nonexistent research papers and authors being misattributed—indicates that AI may have played a larger role in generating certain parts of the content than originally stated.
Several of the cited papers cannot be located in standard academic databases, and some researchers listed as authors confirmed that they had no involvement with the papers attributed to them. These findings point to the potential dangers of relying on AI without rigorous verification, as it can produce convincing but fabricated information that may be mistaken for credible research.
Deloitte said it is currently correcting the errors in citations while stressing that the overall findings, analyses, and recommendations of the report remain accurate and unaffected. The firm emphasized that the report’s conclusions are still valid for guiding government decision-making despite the issues identified in the references.
Government response
The Canadian provincial government paid for the report in eight instalments, totalling nearly $1.6 million. So far, the government has not publicly addressed the issues found in the report. The report remains available on the government’s official website.
The study was prepared for the previous government, which was led by the Liberal Party. Since then, a new leadership team has taken office, but no response has been issued regarding the report’s accuracy or any possible refund.
The report was meant to inform decisions on critical health care policies in a province already facing shortages of nurses and doctors. Errors in such a document could affect the planning and implementation of health services.
Similar issues in other countries
This is not the first time Deloitte has faced criticism for using AI in government reports. Earlier this year, Deloitte produced a report for the Australian government costing $290,000. That report also included non-existent research papers and fabricated quotes.
In Australia, Deloitte later admitted to using AI to create parts of the report and uploaded a corrected version to the government website. Deloitte also had to pay a partial refund for the errors.
In Canada, Deloitte maintains that the main content and recommendations of the report are still accurate, even where minor citation errors occurred. No information has been released about whether a refund or further corrections will be made.
The discovery of AI-related errors in such a high-cost report raises questions about how consulting firms prepare research for government clients. The mistakes show the importance of thorough fact-checking and careful review before reports are used to guide policy decisions.