KPMG Withdrew Its Artificial Intelligence Report. Why? The AI Wrote It

KPMG, one of the Big Four accounting firms, quietly removed a report they had released titled ‘Redefining excellence in the age of agentic AI.’ Why? Because some of the ‘data’ in the report seemed to be hallucinated by AI. This is the good kind of warning — the kind that makes you go back and check your stuff.

This report came out in October 2025. A research group called GPTZero analyzed it and found several errors, which they attributed to AI hallucinations. The report stated that certain organizations were implementing AI in certain ways, but some of the mentioned organizations (like UBS bank, the UK’s National Health Service, Switzerland’s Federal Railways, and Transport for London) said that the information about them in the report was either false or misleading. If actual organizations need to come forward and deny something they supposedly did, then there’s no choice but to remove it. And so KPMG removed it, pending investigation. This was covered by the Financial Times.

1This isn't a KPMG problem — it's an everyone problem

It’s not an isolated incident either. In the previous month, EY (also a Big Four firm) had to retract a loyalty rewards report that contained fake endnotes and AI hallucinations. Remember, these are the most bureaucratic, double-checking, paid-to-double-check organizations in existence. If fake sources can fool them, then they can surely fool you and me. Even KPMG themselves admitted that this was more of a human error than a technological one: “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.” In other words, the AI wasn’t at fault here. The human was.

So what does this have to do with you if you don’t even know how to open Excel? Well, AI hallucinations are convincing. Fake numbers aren’t printed in red. Fake references are formatted identically to real ones. The problem isn’t with blatantly wrong answers, which are easy to spot. It’s with seemingly accurate details that you don’t bother checking, because everything else seems fine. And that’s how it managed to pass through a team of KPMG reviewers, and how it will pass through your emails, papers, and reports this week.

This is not an anti-AI message. These firms weren’t harmed by AI. They were harmed by blind faith in it. And we can learn from their mistakes, without paying a penny. All it takes is thirty seconds to verify things before hitting ‘send’.