KPMG Pulls Report on AI Usage After Apparent Hallucinations

At a glance:

• KPMG removed its AI usage report after multiple organizations disputed its claims
• GPTZero identified inaccuracies linked to AI hallucinations in the October 2025 document
• UBS, NHS, Swiss Railways, and TfL all called the report's assertions misleading or false

What Happened

KPMG's report, titled "Redefining excellence in the age of agentic AI," was published in October 2025 but quickly became controversial. Research group GPTZero analyzed the document and found multiple factual errors, which it attributed to AI hallucinations. The firm's spokesperson confirmed the removal, stating they were investigating the issue while emphasizing their commitment to responsible AI use with human oversight. This isn't an isolated incident—last month, EY also retracted a report on loyalty programs after similar hallucination problems.

The report claimed specific AI implementations by major organizations, but UBS, the UK's National Health Service, Swiss Federal Railways, and Transport for London all denied the accuracy of these claims. GPTZero's analysis revealed patterns consistent with generative AI errors, suggesting KPMG may have relied on automated tools without sufficient human validation. The spokesperson acknowledged this, stating: "We expect all our people to follow our guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources."

Why It Matters

AI hallucinations pose a critical risk to professional reporting, especially when used to generate authoritative documents. KPMG's case highlights how even well-intentioned AI-assisted work can produce misleading information if not properly audited. The involvement of high-profile clients like UBS and NHS amplifies the stakes, as inaccurate data could impact business decisions or public services. This incident also raises questions about the reliability of AI-generated content in regulated industries where precision is non-negotiable.

The broader tech community is watching how KPMG addresses this. While the firm hasn't disclosed technical details about the hallucinations, the EY parallel suggests a growing awareness of AI's limitations in professional services. Organizations using AI for report generation may need to implement stricter validation protocols, potentially slowing down workflows but ensuring accuracy. The incident also underscores the need for clearer industry standards around AI-assisted content creation.

Broader Implications

This event reflects a larger trend where AI tools are being integrated into knowledge work at unprecedented speed. While AI can accelerate research and drafting, KPMG's experience shows the dangers of over-reliance without safeguards. The report's removal signals a potential shift in how professional firms approach AI adoption—prioritizing verification over speed. For clients, this means increased scrutiny of AI-generated outputs, which could lead to higher costs or delays in service delivery.

The healthcare and transportation sectors mentioned in the report are particularly sensitive to data accuracy. A hallucinated claim about AI usage in NHS operations could have real-world consequences for patient care or infrastructure planning. Similarly, UBS's denial of the report's assertions about its AI implementations highlights the competitive risks of misinformation in finance. These sectors may now demand more rigorous AI audit processes before adopting similar technologies.

What to Watch Next

KPMG's internal investigation will be critical to understanding how the hallucinations occurred. The firm may face pressure to disclose whether specific AI models or tools were responsible, and whether this was an isolated case or part of a pattern. The EY incident suggests this isn't unique to KPMG, and other professional services firms using AI for reporting could face similar scrutiny. Regulatory bodies might also take interest, especially given the involvement of public sector clients like the NHS.

The long-term impact depends on how KPMG responds. If they implement stricter AI governance frameworks, it could set a precedent for the industry. Conversely, if they downplay the issue, it might erode trust in their AI capabilities. Clients and competitors will likely monitor their next moves closely, particularly regarding future AI-assisted reporting projects. This case also serves as a cautionary tale for organizations considering AI tools without robust validation mechanisms.

Conclusion

KPMG's situation illustrates the double-edged sword of AI in professional services. While it offers efficiency gains, the risk of hallucinations demands careful management. The incident serves as a reminder that AI should augment—not replace—human expertise, especially in high-stakes environments. As AI becomes more integrated into knowledge work, the balance between innovation and accuracy will remain a central challenge for firms across sectors.