KPMG has pulled its October 2025 report on agentic AI after entities including UBS and the UK’s NHS refuted claims about their AI usage featured in the report, exposing issues of AI hallucination and inadequate human oversight.
- KPMG’s AI report contained fabricated claims about major organizations’ AI use.
- AI hallucinations identified as the cause of the errors by detection firm GPTZero.
- Incident highlights risks of AI-generated content without rigorous human review.
What happened
KPMG published a report in October 2025 exploring AI adoption titled "Redefining excellence in the age of agentic AI." The report included descriptions of AI use by prominent organizations such as UBS, the UK’s National Health Service, Swiss Federal Railways, and Transport for London. However, these entities publicly disputed the report’s claims, stating they were inaccurate or misleading.
Following the disputes, KPMG removed the report from its website and announced an internal investigation. An AI detection company, GPTZero, identified that the report’s errors stemmed from AI hallucinations, meaning the AI-generated fabrications that were not verified by humans before publication.
Why it matters
This incident exemplifies the risks of relying on AI to generate authoritative reports without sufficient human oversight. Professional services firms like KPMG are increasingly embedding AI tools in their workflows, raising concerns about the accuracy and trustworthiness of their public and client-facing outputs.
The KPMG case is especially notable because the report focused on AI adoption itself, yet failed to detect AI-generated inaccuracies. This undermines confidence in the firm’s ability to manage AI responsibly, possibly affecting client trust and raising questions about the quality control processes applied to AI-assisted deliverables.
What to watch next
KPMG’s investigation and subsequent actions will be closely monitored to see how the firm reforms its AI governance practices. It will be important to assess whether enhanced human review and validation measures are implemented to prevent future hallucinations in AI-assisted content.
The broader professional services sector is also under scrutiny as similar incidents have occurred at firms like EY and in national policy documents. This underscores a growing need for standardizing responsible AI use, training, and verification protocols to maintain credibility as AI is more deeply integrated in professional outputs.