KPMG Withdraws AI Report After GPTZero Identifies Hallucinations From AI-Generated Content
Professional services firm pulls October 2025 report on agentic AI after organizations dispute accuracy of AI-usage claims.
Last verified:
KPMG Report Withdrawal Signals Validation Gap in AI-Assisted Research
KPMG, a major professional services firm, has withdrawn its October 2025 report titled “Redefining excellence in the age of agentic AI” after GPTZero identified numerous inaccuracies traced to AI hallucinations embedded in the text. According to TechCrunch AI, four organizations—UBS, the UK’s National Health Service (NHS), Swiss Federal Railways, and Transport for London—independently disputed specific claims about their AI usage reported in the document, confirming that the errors were not isolated instances but systematic misrepresentations.
How Hallucinations Contaminated the Report
The irony of KPMG’s situation lies in its method: the firm appears to have relied on AI tools to generate or assist in writing a report about AI capabilities and adoption, without implementing sufficient human-led verification at critical stages. GPTZero’s analysis, cited by the Financial Times, traced the inaccuracies directly to AI hallucinations—fabricated details that models generate when they lack grounding or when they extrapolate beyond training data. A KPMG spokesperson acknowledged the lapse in governance, stating that the firm expects “all our people to follow our guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources.” The withdrawal signals that this human-in-the-loop process failed to catch the errors before publication.
Broader Pattern in Consulting and Enterprise Research
KPMG’s pullback is not an isolated incident. According to TechCrunch, EY (Ernst & Young), another Big Four consulting firm, withdrew its own report on loyalty rewards programs in May 2026 after discovering that it contained fabricated footnotes and AI-generated hallucinations. The back-to-back withdrawals suggest that major professional services firms are struggling to reconcile rapid AI adoption for content generation with the due diligence standards their clients expect—especially when the subject matter itself involves AI.
Why This Matters
The incident exposes a critical vulnerability in enterprise AI workflows: organizations confident enough to deploy large language models for high-stakes research and client-facing reports may underestimate the validation burden. For teams evaluating third-party AI research or consulting reports, this reinforces the need to independently verify claims, particularly those citing specific organizational practices or benchmark data. For KPMG, EY, and peer firms, the withdrawals will likely prompt stricter content-approval checkpoints and clearer accountability for AI-assisted outputs. For end clients commissioning AI-heavy research, the pattern suggests that “human oversight” language in vendor proposals requires teeth—auditable review logs, named validators, and explicit sign-off on factual claims rather than passive compliance gestures.
Frequently Asked Questions
What inaccuracies did GPTZero find in the KPMG report?
GPTZero identified multiple AI hallucinations in the text. UBS, the UK's National Health Service, Swiss Federal Railways, and Transport for London each stated that claims about their specific AI usage were either untrue or misleading.
Why did KPMG pull the report from its websites?
According to TechCrunch, KPMG removed the report to conduct its own internal investigation into the accuracy issues after external organizations disputed the claims.
Is this the first time a major consulting firm has withdrawn a report due to AI-generated errors?
No. In May 2026, EY withdrew a separate report on loyalty rewards programs that appeared to contain fabricated footnotes and AI hallucinations, suggesting a broader industry pattern.