4 min learnNew DelhiJun 14, 2026 12:06 PM IST
KPMG has withdrawn a research on agentic AI that reportedly included AI-generated hallucinations and faux citations, within the newest instance of knowledgeable providers agency being caught by the shortcomings of the know-how.
The report titled ‘Whole Expertise: Redefining Excellence within the Age of Agentic AI’ was revealed in October 2025. Nevertheless, analysis group GPTZero just lately recognized a lot of inaccuracies within the report, a lot of which stemmed from AI-generated hallucinations. Because of this the corporate appears to have used AI instruments to draft the report about agentic AI.
KPMG pulled the report after a number of organisations talked about within the doc stated that its claims about their AI utilization have been unfaithful. UBS, the UK’s Nationwide Well being Service, Swiss Federal Railways, and Transport for London stated that the KPMG report’s claims about their AI utilization have been both unfaithful or deceptive, in response to Monetary Instances.
The incident underscores how consulting corporations have develop into significantly weak to AI-generated errors, whilst they warning shoppers about safeguarding in opposition to AI hallucinations and different reliability points. It additionally comes only one month after EY withdrew a report on loyalty rewards packages that appeared to incorporate pretend footnotes and AI hallucinations.
Final 12 months, Deloitte was compelled to refund the Australian authorities after AI-generated content material made its approach right into a taxpayer-funded report.
“KPMG Worldwide takes the accuracy and integrity of its revealed content material severely. The report has been eliminated and we’re reviewing the circumstances surrounding its publication. We count on all our folks to observe our tips on the accountable use of AI, together with human oversight to validate content material and confirm unbiased sources,” an organization spokesperson was quoted as saying by The Register.
What did researchers discover?
Upon conducting a forensic overview of KPMG’s report, GPTZero researchers stated that they discovered solely 5 out of 45 citations within the report accurately pointed to the cited supply. The remaining citations have been both deceptive, partially fabricated, or too obscure to confirm. The pattern of generative AI instruments making up citations by stitching collectively items of actual sources or inventing completely new ones has been dubbed as ‘vibe citing’ by GPTZero.
Story continues beneath this advert
Moreover, almost half of the report’s claims have been false, unsupported, or attributed to the unsuitable supply. A number of research talked about within the report that seemed to spotlight cutting-edge deployments of AI brokers have been additionally discovered to be artistic, as per the researchers.
Case research about purported agentic AI deployments at organisations comparable to UBS, Swiss Federal Railways, and Transport for London both did not substantiate the report’s claims or contained alterations and paraphrasing that undermined their reliability, they stated.
“These factual errors usually are not confined to the report’s footnoted passages. On web page 42, the authors declare that Emirates airline has adopted a cell chatbot named Sara (false) that may converse instantly with passengers (partially true) and alter their flights (false). In reality, Sara is a robotic assistant launched by Emirates in 2023 (not a chatbot) that lacks the flexibility to change flight bookings,” GPTZero researchers stated.
The analysis agency additional famous that sure figures appeared to contradict the agency’s earlier studies. As an illustration, a statistic that stated over 55 % of CEOs ranked AI as their high funding precedence was in contradiction with the 71 per cent determine talked about in KPMG’s 2025 CEO Outlook that got here out throughout the identical month.