A report on the future of AI, published by global consulting giant KPMG, was found to contain approximately 40 AI hallucinations out of just 45 citations before its withdrawal. KPMG aimed to define excellence in agentic AI, but its own report demonstrated a profound lack of accuracy due to widespread AI-generated errors. This incident demands rigorous human oversight in AI-driven research, even from trusted sources, or risk significant reputational damage and widespread misinformation.
The Extent of the Problem: Dozens of Hallucinations
KPMG's report on AI's future contained approximately 40 AI hallucinations out of 45 citations, as reported by PCMag. Released in October 2025 (likely a typo given its 2026 withdrawal, and now outdated), the report's inaccuracies were identified by GPTZero, attributing errors to AI hallucinations, according to IndexBox. This sheer volume of errors reveals a systemic failure in KPMG's verification process for an AI-focused publication. Such inaccuracy suggests a dangerous over-reliance on unverified outputs, fundamentally eroding client trust in 'expert' advice.
Organizations Dispute False AI Adoption Claims
UBS, the UK’s National Health Service, Swiss Federal Railways, and Transport for London all disputed KPMG’s claims about their AI usage, labeling them untrue or misleading, reports TechCrunch and The Indian Express. These public refutations forced KPMG's hand, escalating reputational damage beyond internal control. Such high-profile misattributions show AI hallucinations can fabricate entire use cases, damaging the credibility of both the report and the cited organizations. This incident exposes a critical vulnerability for professional services firms whose brand reputation relies on trust and expertise.
Fabricated Use Cases and Outdated References
The KPMG report falsely claimed Japanese East Japan Railway Company (JR East) uses agentic AI for customer service, linking to an outdated 2019 press release, PCMag reported. This specific fabrication, tied to an outdated press release, exposes AI's ability to generate plausible but false connections, undermining factual integrity. The nature of these hallucinations—attributing advanced AI usage to identifiable corporations—proves AI models can generate highly specific, fabricated claims. This makes general fact-checking insufficient, demanding direct verification from named entities.
Misleading Claims About Advanced AI Adoption
The report also falsely claimed Austrian electricity provider Verbund uses AI agents in households for real-time analytics, PCMag noted. Such detailed but false claims about advanced AI adoption risk misleading other companies or policymakers about agentic AI's true state and readiness. Firms failing to implement robust validation processes for AI-generated content not only publish misinformation but actively undermine the entire AI industry's credibility. This proves its current unreliability in even basic fact-checking.
The KPMG debacle is a powerful wake-up call: the widespread adoption of agentic AI will likely hinge on the industry's commitment to rigorous human oversight, as unverified outputs could otherwise trigger a profound crisis of trust.
