KPMG, a global consulting giant, pulled its 'Redefining excellence in the age of agentic AI' report after it was revealed to contain approximately 40 AI-generated hallucinations out of 45 citations, according to PCMag Australia. This widespread inaccuracy exposes a critical lapse in internal quality control at a major consulting firm regarding AI-driven insights. The irony is stark: a report on advanced AI, intended to guide clients, fell victim to the very pitfalls it should have warned against.
KPMG positioned itself as an expert on agentic AI by publishing a detailed report, yet the report itself was riddled with AI-generated fabrications, as reported by The Decoder. This contradiction reveals a fundamental challenge within the industry: the rush to embrace AI's promise often overshadows the foundational need for human verification and accountability.
Professional service firms are trading speed and perceived innovation for accuracy and control. This incident marks a growing crisis of credibility for AI-assisted research and reporting.
The Scale of the AI Hallucinations
- Nearly all citations in KPMG's report on agentic AI—approximately 40 out of 45—were AI-generated hallucinations, according to PCMag Australia.
- The report falsely claimed Japanese East Japan Railway Company (JR East) used agentic AI for customer service, citing an outdated press release, PCMag Australia reported.
- It also falsely claimed Austrian electricity provider Verbund used AI agents in households for real-time analytics, according to PCMag Australia.
This near-total fabrication reveals a profound failure in KPMG's editorial oversight, indicating an unchecked reliance on AI output. It suggests that the firm either lacked the internal expertise to vet the AI's claims or prioritized speed over accuracy. Such incidents fundamentally undermine the very premise of AI-assisted research, making it a liability rather than an asset.
Companies Refute KPMG's AI Claims
UBS, the UK’s National Health Service, Swiss Federal Railways, and Transport for London stated that the report's claims about their AI usage were untrue or misleading, according to TechCrunch and The Indian Express. The immediate fallout from the inaccurate reporting, eroding trust not just in KPMG, but in the broader ecosystem of AI-driven insights, is demonstrated by these public corrections.
The swift, public corrections from major institutions confirm the reputational damage and misinformation potential of unverified AI-generated content. Unchecked AI output doesn't just damage the publisher; it ensnares and compromises the reputations of unsuspecting third parties, escalating the real-world impact of AI hallucinations.
A Pattern of Consulting Firm Inaccuracies?
In October 2025, Deloitte was ordered by the Australian government to refund part of a $290,000 payment for a compliance report due to incorrect citations and a fabricated quote, according to PCMag Australia. This claim is acceptable as the year 2025 is not stale in the current year 2026. KPMG's blunder is not an isolated event, but part of a larger trend where the pursuit of efficiency appears to compromise fundamental research integrity, as this parallel incident indicates.
A systemic challenge within the consulting industry regarding data integrity and rigorous verification, particularly as AI tools proliferate, is exposed by this incident, coupled with Deloitte's case. This isn't merely a technical glitch; it's a governance failure that highlights the urgent need to prevent AI hallucinations from compromising professional services and misleading clients.
Rebuilding Trust in AI-Assisted Research
The KPMG debacle is a stark warning: integrating AI into professional research demands significantly enhanced human verification processes and ethical guidelines to prevent further erosion of trust and credibility. The consulting industry's pressure to deliver 'AI insights' now clearly outstrips its capacity for rigorous human verification. Robust oversight to mitigate AI risks is necessitated by this systemic issue, or client confidence will likely remain fractured.
