KPMG Pulls AI Usage Report After Apparent Hallucinations

KPMG has withdrawn an AI report after several claims in the document appeared to be inaccurate or based on AI-generated hallucinations.

The report, titled “Redefining excellence in the age of agentic AI,” was published in October 2025 and focused on how major organizations were using agentic AI. However, research group GPTZero and the Financial Times found that some of the report’s case studies and citations did not match reality.

The incident is especially notable because the report was about the benefits of AI, but it now appears to have included the same type of AI errors it was discussing.

What Happened?

KPMG removed the report after multiple organizations named in it disputed the way their AI usage was described.

The report reportedly included examples involving companies and public bodies such as UBS, the UK’s National Health Service, Swiss Federal Railways and Transport for London. According to reports, some of these organizations said the claims were wrong, misleading or did not reflect their actual AI use.

GPTZero said many of the report’s citations appeared to be unreliable. The Register reported that only 5 of the report’s 45 citations matched their sources correctly, while other citations were either inaccurate, distorted or difficult to verify.

Why the Report Was Pulled

The main issue was not just small factual errors. The concern was that the report may have used AI-generated references or summaries that sounded credible but were not accurate.

This type of error is often called an AI hallucination. It happens when an AI system produces information that looks real but is false, unsupported or wrongly attributed.

In this case, the problem was serious because KPMG is one of the world’s largest professional services firms. Reports from major consulting companies are often used by executives, journalists, policymakers and businesses to understand market trends.

If such reports contain false examples or weak citations, they can spread misinformation quickly.

Why This Is Embarrassing for KPMG

The controversy is awkward for KPMG because the report was focused on AI adoption and the promise of agentic AI.

Agentic AI refers to AI systems that can perform tasks with more autonomy, such as planning, making decisions and completing workflows with less human input.

KPMG and other major consulting firms have been advising companies on AI strategy, governance and transformation. Because of that, clients expect a high level of accuracy and fact-checking from their research.

A report about AI containing apparent AI hallucinations creates a trust problem. It raises questions about how the report was produced, reviewed and approved before publication.

KPMG Opens an Internal Review

KPMG has reportedly started an internal investigation into how the report was created and why the inaccuracies were not caught earlier.

The company has also removed the report from multiple websites. This suggests that KPMG is trying to limit the spread of the disputed claims while it reviews the issue.

The case may push consulting firms to strengthen their internal rules around AI-assisted research, especially when publishing public reports that include case studies, citations and client-related examples.

Bigger Problem: AI Errors in Professional Reports

This is not the first time a major professional services firm has faced problems with AI-generated errors.

EY recently retracted a cybersecurity-related report after researchers found fabricated citations and other AI-style inaccuracies. Similar issues have also appeared in legal filings, academic papers and corporate documents.

The pattern shows a growing risk: AI can help teams write faster, but it can also create convincing false information if humans do not verify every claim.

This is especially risky in industries where credibility matters, such as consulting, law, finance, healthcare and government policy.

Why It Matters

The KPMG case highlights a major challenge for businesses using generative AI.

Many companies are encouraging employees to use AI to improve productivity. But if AI-generated content is not carefully checked, it can damage trust, create legal risk and spread false information.

The issue is not that AI cannot be useful. The issue is that AI outputs need strong human review, source checking and editorial control.

For companies publishing research, reports or market analysis, this means every citation, statistic, quote and case study must be verified before release.

Final Take

KPMG pulling its AI report is a reminder that AI can create serious credibility risks when used without enough oversight.

The irony is clear: a report promoting AI appears to have suffered from one of AI’s most well-known weaknesses.

As more companies use generative AI for research, writing and analysis, the pressure will grow for stronger fact-checking systems. AI may speed up content creation, but trust still depends on human verification.