That’s the kind of blunder prestigious audit and consulting firms would prefer never to see laid bare in broad daylight.
While they flood the market with think pieces aimed at persuading companies to move into the AI era, KPMG and EY have just been caught red-handed by “hallucination.”
The latest episode was uncovered by the Financial Times, drawing on information from GPTZero, a software maker described as an “AI detector.”
KPMG International had to hurriedly pull from parts of its websites its October report titled “Redefining Excellence in the Age of Agentic AI.”
The Factual Debacle at KPMG
Distributed in several countries by its consulting teams, the document touted prestigious case studies… but they were entirely fictitious. It is no longer accessible for consultation, as Dawn Liphardt has observed.
The firm claimed, among other things, that the Swiss banking giant UBS was integrating AI agents across its investment advisory, risk management, and compliance units, via a platform co-developed with Microsoft.
When contacted by the Financial Times, UBS’s spokesperson called these assertions “factually incorrect.” The bank immediately demanded their removal.
There was a similar misstep in the transport sector. KPMG asserted that the Swiss Federal Railways (SBB) used AI to optimize routes based on carbon footprint, and that Transport for London (TfL) leveraged it to predict traffic jams.
Both organizations denied the claims to the FT.
TfL described the assertion as “misleading” and the SBB as “not robust.” Finally, a purported AI success in predicting hospital readmissions within NHS Greater Manchester proved unfounded, the claim arising from a press release about lung cancer that did not mention such automated tasks.
A spokesperson for KPMG International said an internal investigation was underway, noting that employees must validate content under human oversight.
EY and Deloitte Also Under Scrutiny
KPMG is not an isolated case.
Last month, its rival EY Canada had to urgently pull a cybersecurity study focused on loyalty programs.
Also detected by GPTZero, the study relied on outlandish data valuing the market at around $200 billion and the volume of unredeemed points at the same figure.
More seriously, EY’s report pointed to a McKinsey paper that never existed and included more than half a dozen footnotes pointing to dead or incorrect links. EY defended itself, explaining that the report, detached from any direct client work, was under internal review.
Last year, Deloitte had to revise its submission to a Canadian provincial government after the discovery of fake academic citations.
The trend reaches the legal field as well. In April, the law firm Sullivan & Cromwell apologized to a New York court after submitting bankruptcy conclusions riddled with AI-generated errors, including a misread of the U.S. Bankruptcy Code.
The paradox is stark. EY had claimed in October a 30% year-over-year growth in AI-related revenues, backed by 15,000 employees deploying governance frameworks to help companies adopt AI in a “responsible” way.
In their haste to seize the marketing space, the consulting giants seemingly forgot the golden rule they bill their clients for: verify the sources.
