KPMG just pulled one of its flagship AI reports after researchers found it was built almost entirely on fake citations, and other outlets had already cited it before the takedown. KPMG’s 2025 global AI trust study flagged this exact risk, finding that 56% of AI users made errors from hallucinations. The firm warned clients about this for over a year, then published a report hitting the same problem. This stings more because of who’s involved. KPMG isn’t a startup figuring out AI. It’s one of the four largest professional services firms globally, the kind whose reports get treated as fact by default. Yet fabricated citations still made it to publication.Four organizations publicly denied the AI success stories KPMG attributed to them. Do you think that AI is so random and it cannot be trusted ?
