KPMG pulls AI report after claims prove inaccurate

KPMG’s latest AI message has turned into a governance problem for the firm’s own people. After claims in a report on agentic AI were challenged by several organizations, the episode exposed how much pressure now sits on review standards, source checking, and human signoff inside a professional-services business that sells trust for a living.

KPMG pulled Redefining excellence in the age of agentic AI after UBS, the NHS, Swiss Federal Railways and Transport for London disputed claims attributed to them. GPTZero said its review found inaccuracies in the report and blamed hallucinations, and said only five of the report’s 45 citations correctly matched their sources. KPMG removed the report from its websites while it investigated, and a spokesperson said staff are expected to follow responsible AI guidelines that require human oversight to validate content and verify independent sources.

For KPMG teams, the fallout reaches beyond one embarrassing PDF. The same controls that are supposed to protect client work now look central to internal publishing, too, especially in proposal writing, industry thought leadership, benchmarking materials and any workflow where generative AI helps draft or research content. In a firm built on audit discipline and advisory credibility, a report with disputed citations is not just a communications issue. It raises the harder question of whether editorial, research and approval steps are strong enough when AI speeds up the first draft faster than people can verify the facts.

That question lands even more sharply because KPMG has been publicly leaning into AI. In April 2026, its Trusted AI framework said the firm is committed to using AI ethically and responsibly, with a cross-functional Trusted AI council that continually reviews the framework and principles meant to guide AI systems across the lifecycle. KPMG’s 2025-2026 customer experience report also frames its annual research around the Customer Experience Excellence score and the Six Pillars of experience, which makes the disputed agentic AI report feel more like a research-quality failure than a stray marketing error.

The reputational risk is not limited to one report. GPTZero said this was its second investigation in a series focused on hallucinated citations, and the episode echoes an earlier consulting-industry case in which Deloitte refunded the Australian government after AI-generated content slipped into a taxpayer-funded report. KPMG’s own January 2026 global tech report, based on a survey of 2,500 tech executives from 27 countries, shows how aggressively the firm is still pitching AI transformation across audit, tax and advisory. The lesson for KPMG employees is straightforward: the firm can sell AI only if its own review process is slower, tougher and more traceable than the tools it is deploying.