KPMG AI Report Hallucinations — UBS NHS TfL — Finance Monthly

KPMG has withdrawn a serious report on synthetic intelligence adoption after an investigation discovered it contained fabricated case research attributed to UBS, NHS Higher Manchester, Transport for London and Swiss Federal Railways — errors recognized as AI hallucinations by analysis group GPTZero and subsequently verified by the Monetary Occasions.

The October report, titled “Redefining excellence within the age of agentic AI”, claimed that UBS integrates AI brokers throughout funding advisory, threat administration and compliance monitoring by way of a platform co-developed with Microsoft. A spokesperson for the Swiss financial institution instructed the FT the assertions have been factually incorrect.

The report made equally false claims about Swiss Federal Railways, stating that the transit group makes use of AI brokers to plan and optimise passenger journeys — a characterisation the railway confirmed was not correct. Transport for London mentioned a declare that it makes use of AI brokers to foretell and handle congestion was deceptive, whereas NHS Higher Manchester mentioned assertions about AI-driven affected person triage and hospital readmission prediction didn’t align with the press launch the footnotes indicated as their supply.

KPMG Worldwide has since eliminated the report from its web sites whereas it investigates the circumstances surrounding its publication. A spokesman mentioned the agency expects all workers to comply with pointers on accountable AI use, together with human oversight to validate content material and confirm impartial sources.

The episode will not be remoted. EY retracted a examine final month after GPTZero recognized pretend footnotes and different errors, and Sullivan & Cromwell admitted in April {that a} chapter courtroom submitting contained AI-generated inaccuracies together with misreadings of the US chapter code.

GPTZero, the AI detection analysis group that recognized the errors and whose findings have been verified by the Monetary Occasions, has documented a rising sample of hallucinated content material throughout skilled providers publications. The credibility connected to corporations reminiscent of KPMG and EY means fabricated findings flow into extensively earlier than corrections are issued — the KPMG report had already been referenced by a number of trade publications and a serious Czech newspaper earlier than its withdrawal, amplifying the governance threat for any organisation that acted on its findings.

For CFOs and finance administrators who’ve relied on Massive 4 thought management to benchmark AI adoption or develop inner governance frameworks, the implications are rapid. The analysis underpinning strategic know-how selections at board stage is demonstrably fallible when produced at quantity with out ample human evaluate.

Companies which have cited KPMG’s findings in their very own AI technique paperwork face the duty of revalidating these references. Extra broadly, the episode reinforces the case for finance leaders to demand major supply verification earlier than performing on externally produced AI analysis — whatever the repute of the agency that produced it.