KPMG withdraws AI report: 40 of 45 sources were fabricated - second consulting scandal from AI hallucinations in one month
What it really says
KPMG has withdrawn its report 'Redefining Excellence in the Age of Agentic AI' after multiple organizations named in the study challenged the claims as false. Originally published in October 2025, the report described in detail how prominent organizations were allegedly deploying AI agents in their business operations. AI detection platform GPTZero examined the report and found that of 45 cited sources, only five were accurately linked - the remaining 40 were either completely fabricated or misleading. The Financial Times independently verified these findings. Specifically: the report claimed UBS operates 'AI agents for investment advice, risk management and compliance monitoring on a platform jointly developed with Microsoft' - a UBS spokesperson stated this was 'not true.' NHS Greater Manchester was portrayed as a pioneer in AI-powered patient management, automated referrals, and prediction of hospital readmissions - representatives from the healthcare body said these claims did not align with reality. Swiss Federal Railways (SBB) and Transport for London also denied the report's claims. GPTZero told the Financial Times that the errors stemmed from AI hallucinations. This is the second such incident within a month: consulting firm EY had previously withdrawn a study after GPTZero similarly identified fabricated references and AI-generated errors.
Our assessment
This case is significant for the AI debate on multiple levels. First, it reveals a real problem: when one of the world's four largest accounting firms apparently uses AI to produce a report and fails to detect hallucinated content, it raises fundamental questions about quality assurance across the entire consulting industry. KPMG is not a small startup - it advises governments and major corporations. Second, a troubling pattern is emerging: two Big Four-adjacent scandals within a month (EY and KPMG) suggest the problem may be systemic. Third, there is also a reassuring side: the errors were detected - by GPTZero (an AI detection platform) and the Financial Times. Self-correction works. The report was withdrawn, the companies pushed back. Tools and processes exist that can uncover hallucinated content. The real danger is not that AI hallucinates - that is well known - but that people accept AI-generated content as fact without verification. The KPMG case is ultimately a warning signal for anyone publishing AI-generated text without human review.
Relevance for Germany
This case is directly relevant for Germany. KPMG is one of the leading auditors and management consultants in Germany, advising numerous DAX corporations, mid-sized companies, and the public sector. If KPMG reports can contain AI-generated misinformation, it raises the question of how reliable other studies and analyses are that influence business decisions in Germany. The EU AI Act, which introduces transparency obligations for AI-generated content from August 2026, could help in the long term: when it must be clearly labeled where AI has contributed to writing, pressure on human quality control also increases. For German companies using AI for report creation, this case is a clear warning: without rigorous review by subject matter experts, AI hallucinations can end up in official documents. The TUV Association and the BSI have repeatedly highlighted the need for AI quality standards - this case underscores their demands.
Fact check
Core information is confirmed by multiple independent sources. TechCrunch reported on June 13, 2026 about the report withdrawal, naming GPTZero as the discoverer of errors. The Register reported as early as June 12. The Financial Times independently verified GPTZero's findings. German media (t3n, The Decoder, IT-Boltwise) report consistently. The figure of 40 erroneous sources out of 45 comes from GPTZero's analysis. The denials from UBS, NHS Greater Manchester, SBB, and Transport for London are documented through the Financial Times investigation. The EY precedent from the previous month is also multiply documented. The title of the KPMG report ('Redefining Excellence in the Age of Agentic AI') and the original publication date (October 2025) are consistently reported.
Source
- • https://techcrunch.com/2026/06/13/kpmg-pulls-report-on-ai-usage-due-to-apparent-hallucinations/
- • https://www.theregister.com/ai-and-ml/2026/06/12/kpmgs-ai-report-turns-into-a-demo-of-ai-hallucinations/5255029
- • https://gptzero.me/news/investigations-kpmg/
- • https://the-decoder.de/blamage-fuer-kpmg-bericht-ueber-ki-in-unternehmen-enthielt-erfundene-fallstudien/
- • https://t3n.de/news/wegen-halluzinierter-fallstudien-kpmg-zieht-studie-zur-nutzung-von-ki-agenten-zurueck-1747573/