Another Deloitte report accused of AI-made citations, this time in Canada

Fresh allegations of AI-made citations in a Canadian report deepen questions over Deloitte's evidence checks, weeks after similar issues surfaced in Australia

Deloitte, one of the world’s most influential consulting firms, is contending with a pair of AI-linked embarrassments: first in Australia, now in Canada. 

 

What began as a standard consulting assignment has snowballed into a reputational crisis for Deloitte. The firm insists it uses AI selectively. But two separate investigations, first in Australia and now in Canada, have raised doubts about whether those tools are slipping into its work in ways the firm cannot fully control. After the Australia fiasco, and now in Canada too, researchers cited in one of its reports say they were linked to studies that don’t exist.

 

An investigation into a major health workforce report prepared for the government of Newfoundland and Labrador by Deloitte has revealed that parts of the document appear to rely on inaccurate or nonexistent academic citations. The issues were identified by a Canadian news outlet The Independent. The study was commissioned by the government at around $1.6 million, which was paid in eight installments, the paper said. It has raised questions about the firm’s use of AI tools and its quality-control practices in publicly funded research.

Errors found in Deloitte report for the Canadian govt 

The report in question is a 526-page Health Human Resources Plan released by the provincial government in May this year, which was commissioned by a previous Liberal administration to help shape policy on recruitment, retention and workforce planning, The Independent said. The province has faced persistent shortages of nurses and physicians, and the report was intended to guide long-term strategy.

 

According to the news outlet’s review, several citations were linked to papers that either do not support the claims made or do not appear to exist. These references were used to justify assertions on topics including financial incentives for health workers, the cost-effectiveness of local recruitment, the role of virtual care and the pandemic’s impact on staff.

Researchers dispute papers attributed to them 

One section of the report cites a published article co-authored by Martha MacLeod to support the idea that monetary recruitment incentives generate cost savings because they are cheaper than hiring and training new staff. MacLeod, a professor emerita at the University of Northern British Columbia’s School of Nursing, was quoted by The Independent that the citation is “false” and “potentially AI-generated”.

 

A second citation at the centre of the investigation refers to a paper the Deloitte report calls “The cost-effectiveness of local recruitment and retention strategies for health workers in Canada.” The paper, according to the Independent, cannot be found. One of the authors named in that reference, Dalhousie University adjunct professor Gail Tomblin Murphy, said the paper “does not exist.”

 

The news outlet also reported that the Deloitte document referred to a paper from the Canadian Journal of Respiratory Therapy that could not be located in the journal’s database.

Deloitte acknowledges corrections but defends findings 

Responding to the allegations, Deloitte Canada said it continues to support the health workforce recommendations contained in the report. “Deloitte Canada firmly stands behind the recommendations put forward in our report,” a spokesperson said, as quoted by The Independent. The firm also confirmed that revisions are under way. “We are revising the report to make a small number of citation corrections, which do not impact the report findings.”

 

The spokesperson said AI “was not used to write the report” but was “selectively used to support a small number of research citations".

Deloitte’s run-in with Australian govt over AI-report 

The concerns in Canada come right after the incident in Australia involving a Deloitte-produced report on welfare compliance. That $290,000 study, published in July, was later found to contain fabricated academic references and even a fictitious judicial quote. A revised version uploaded to the Australian government’s website acknowledged the use of Azure OpenAI in producing the document.

 

Deloitte’s Australian member firm had to provide a partial refund to the federal government for the errors. However, no information has been released in Canada about whether any refund will be considered for the Newfoundland and Labrador report.