KPMG Pulls AI Report After False References and Claims Are Exposed

Global consulting and accounting giant KPMG removed from the web a report on how large organizations are integrating AI agents after it was found to contain incorrect or unverifiable examples and citations. The report, published in October and titled “Reimagining Excellence in the Age of Agentic AI,” drew attention after an evaluation by GPTZero, a company that detects AI-generated content, found that only 5 of 45 reviewed references led to the correct sources.

Several organizations named in the report rejected the claims attributed to them. UBS denied the report’s assertion that it operates an AI-based risk management system and asked for the reference to be removed. London’s transport authority and the Swiss railway company also said the descriptions in the document did not reflect reality.

GPTZero also flagged a claim about Emirates, which the report said ran a chatbot called “Sara” that could talk to passengers and change flight bookings. The researchers said that was wrong, explaining that “Sara” is actually a robotic assistant introduced in 2023 and cannot modify reservations.

The researchers described the phenomenon as “Vibe Citing,” in which an AI model mixes partial information, invents plausible-sounding headlines and sources, and produces references that look credible until checked. KPMG said it had opened an internal investigation and stressed that it is committed to accuracy and reliability. The firm also said its staff must follow responsible AI-use guidelines, including human review and verification against independent sources.