📰 Key Takeaways

KPMG, a major international professional services firm, published a report titled “Redefining Excellence in the Agentic AI Era” in October 2025, but multiple organizations cited in the report denied the accuracy of the descriptions, leading KPMG to ultimately remove the report from its website and launch an internal investigation.

The case was exposed by AI detection research organization GPTZero. GPTZero pointed out to the Financial Times that the report contained multiple fabrications, stemming from AI hallucinations — where AI models generate seemingly plausible content without factual basis. In other words, this report exploring the current state of enterprise AI adoption was likely itself produced with AI assistance but lacked proper human verification.

The organizations named include UBS, the UK’s National Health Service (NHS), Swiss Federal Railways, and Transport for London — all told the Financial Times that the descriptions of their AI usage were either inaccurate or misleading. A KPMG spokesperson said the firm requires all staff to follow internal guidelines on responsible AI use, including mandatory human oversight to verify content and cross-check independent sources.

This isn’t an isolated incident — just last month, another Big Four accounting firm, EY, also withdrew a loyalty rewards program report that was found to contain suspected AI-generated fake citations and hallucinated content. The two incidents in succession reveal a serious gap in quality control for AI-assisted writing at top professional services firms.

💬 JudyAI Lab’s Take

KPMG had to pull a research report specifically about the current state of enterprise AI adoption — and the bitter irony is that the report itself may have been a product of AI hallucinations. This perfectly captures the core trust crisis in AI-assisted writing today.

This isn’t just one firm’s PR crisis — it’s a structural warning for the entire “AI-assisted content” ecosystem. The report cited UBS, NHS, Swiss Federal Railways, and Transport for London, but all four organizations denied the accuracy of the descriptions. In other words, after the AI generated the finished draft, nobody actually verified the quoted sources. For AI builders, the most cautionary takeaway is this: the smoother and more complete the AI output, the easier it is for human reviewers to let their guard down. EY also withdrew a loyalty research report containing fake citations around the same time, and two major firms falling short in succession shows this is a systemic workflow design flaw, not an individual employee’s oversight.

We recommend adding a mandatory checkpoint to all AI-assisted content workflows: “Source quotes must be independently verified” — especially those paragraphs that read the most smoothly and completely are often exactly where the most skepticism is needed.

📅 Source Info

🔗 Further Reading