KPMG pulls AI report over fake citations
KPMG withdrew its 'agentic AI' report after UBS, the NHS, Swiss rail and Transport for London said its AI-written claims about them were false.
KPMG has withdrawn its October 2025 report “Redefining excellence in the age of agentic AI” after four of its showcase organizations said the claims about them were false. UBS, NHS Greater Manchester, Swiss Federal Railways and Transport for London were each described as running AI agents they do not use, and all four told the Financial Times the descriptions were wrong.
The errors were not small. The report said UBS runs AI agents across investment advisory, risk management, and compliance through a platform co-developed with Microsoft. UBS said that was factually incorrect. The report claimed Swiss Federal Railways uses AI agents to plan passenger journeys, which the railway said was not accurate, and that Transport for London uses agents to manage congestion, which TfL called misleading. GPTZero, an AI-detection firm, examined the document and told the FT the fabrications came from AI hallucinations. KPMG removed the report and said it is investigating.
Why this matters
A Big Four firm that sells AI governance shipped a flagship report whose evidence was partly invented by a model, and nobody checked it against the named sources before publishing. This is not a one-off. EY recently retracted a study after GPTZero found fake footnotes, and the law firm Sullivan and Cromwell admitted an April court filing contained AI-generated errors.
If you publish anything an AI helped write, the lesson is the same one behind the verification habit and what AI gets wrong: the output is a draft to verify, not a fact to trust. A German court has already held a company liable for false AI answers.