The Irony of Automation: KPMG Pulls Major AI Report Following Widespread Hallucination Allegations

In a striking illustration of the risks inherent in the unchecked deployment of generative artificial intelligence, professional services giant KPMG has been forced to retract a flagship thought-leadership report. The document, titled “Redefining excellence in the age of agentic AI,” was intended to position the firm at the vanguard of the AI revolution. Instead, it has become a cautionary tale about the perils of relying on the very technology firms are aggressively selling to their clients.

The report, published in October 2025, made a series of bold assertions regarding the AI adoption strategies of major global organizations. However, following inquiries from the Financial Times and evidence presented by the research group GPTZero, it became clear that the document was riddled with fabricated claims. Major institutions—including UBS, the UK’s National Health Service (NHS), Swiss Federal Railways, and Transport for London (TfL)—confirmed that the report’s characterizations of their internal AI initiatives were either entirely false or grossly misleading.

The Chronology of a Corporate Misstep

The failure of “Redefining excellence in the age of agentic AI” did not occur in a vacuum; it is the culmination of a broader trend where firms are racing to publish high-volume, AI-generated content to maintain market visibility.

• October 2025: KPMG releases its report, intended to showcase the firm’s deep expertise in agentic AI—the next frontier of technology where AI systems perform autonomous tasks. The report features detailed case studies of how major corporations are supposedly utilizing these systems.
• Late October/Early November 2025: Researchers at GPTZero, a platform specialized in detecting AI-generated text and potential hallucinations, conduct an analysis of the report. They identify significant discrepancies between the report’s claims and the publicly verifiable records of the companies mentioned.
• November 2025: The Financial Times initiates a formal investigation, reaching out to the companies cited in the report. The feedback is unanimous: the companies had not implemented the AI systems described, nor had they collaborated with KPMG in the manner suggested by the text.
• November 2025 (Immediate Aftermath): Faced with mounting evidence of inaccuracy, KPMG pulls the report from its global digital platforms and initiates an internal investigation into the editorial processes that allowed such fabrications to reach publication.

Supporting Data and the Mechanics of the Failure

At the heart of this controversy is the phenomenon of "hallucination"—a term used to describe instances where AI models, tasked with synthesizing information, generate plausible-sounding but entirely fabricated data.

According to experts at GPTZero, the report displayed the classic hallmarks of AI-assisted drafting gone wrong. Large Language Models (LLMs) are probabilistic engines designed to predict the next token in a sequence; they are not inherently truth-seeking databases. When tasked with writing a technical report about AI, the tool appears to have "hallucinated" successful use cases to fulfill the narrative structure requested by the user.

For example, the report claimed that specific transport and healthcare entities had integrated complex agentic AI workflows. When asked for comment, a spokesperson for Transport for London stated that they had no knowledge of the partnership or the specific technical implementation described. Similarly, the NHS and UBS noted that while they are engaged in various digital transformation projects, the specific claims made by KPMG were factually incorrect.

The failure here suggests a breakdown in the "human-in-the-loop" requirement. While firms often market the benefits of AI to their clients, this incident suggests that the same firms are failing to apply the rigorous oversight they preach to their own internal marketing and communications departments.

Official Responses and Accountability

The fallout from the retraction has been swift, forcing KPMG to address the internal systemic failure. In a statement provided to the media, a KPMG spokesperson emphasized that the firm expects its employees to adhere to strict guidelines.

“We expect all our people to follow our guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources,” the spokesperson said. The firm further confirmed that the report has been scrubbed from its websites while a comprehensive investigation into the authoring process is conducted.

The irony of a professional services firm—which charges a premium for its expertise in governance, risk, and compliance—failing to govern its own publication process has not been lost on industry observers. Critics argue that this incident exposes a "publish-or-perish" culture within major consultancy firms, where the pressure to demonstrate "AI-first" credentials leads to the neglect of basic fact-checking protocols.

The Broader Context: A Pattern of Professional Errors

KPMG is not alone in this struggle. This incident is part of a growing pattern of professional services firms stumbling in their adoption of generative AI.

Just last month, the global accounting and consultancy firm EY (Ernst & Young) was forced to withdraw a high-profile report concerning loyalty rewards programs. In that instance, the report was found to contain fake footnotes and references to non-existent studies—classic indicators of an AI model being asked to provide "citations" for its claims. When the model could not find real data, it simply generated convincing-looking, yet completely fraudulent, citations to complete the task.

These back-to-back failures involving two of the "Big Four" accounting firms suggest a systemic issue. These firms, which act as advisors to governments and the world’s largest corporations on AI adoption, are currently struggling to demonstrate the very competence they sell.

Implications for the Consulting Industry

The implications of these errors are profound and extend beyond mere reputational damage.

1. The Erosion of Trust

Professional services firms operate on the currency of trust. Clients hire these firms precisely because they assume the information provided is vetted, accurate, and authoritative. When a firm publishes AI-hallucinated data, it undermines its role as a trusted advisor. If the firm cannot verify its own research, how can it be trusted to manage the complex digital transformations of its clients?

2. The Legal and Compliance Risks

If firms are using AI to draft content that makes claims about third parties, they risk potential litigation. Defamation, trade libel, or misleading corporate disclosures could arise if an AI-generated report falsely claims that a publicly traded company is utilizing a specific, perhaps unproven or experimental, technology.

3. The Need for "Human-in-the-Loop" Reinforcement

The current failures underscore the danger of "automation bias"—the human tendency to favor suggestions from automated systems even when those suggestions are clearly wrong. Employees at these firms likely used AI to draft the reports and, assuming the technology was "smart" enough to be accurate, failed to conduct the necessary due diligence. Moving forward, firms will need to implement mandatory, audited fact-checking workflows that treat AI-generated content with the same skepticism as unverified web content.

4. A Shift in Strategy

This incident may lead to a cooling-off period in the aggressive push for "AI-generated everything." We are likely to see firms adopt more restrictive policies, requiring clear disclosures when AI is used in the creation of corporate literature, and perhaps even a return to human-authored analysis as a "premium" offering to reassure clients of the accuracy of the output.

Conclusion: The Double-Edged Sword of AI

The KPMG and EY incidents serve as a vital reality check for the corporate world. Generative AI is a powerful tool for productivity, capable of summarizing documents, drafting emails, and organizing vast amounts of data. However, it is fundamentally ill-equipped to act as an objective, fact-based researcher without intense, expert human intervention.

As the industry moves forward, the "age of agentic AI" must be balanced with the age of accountability. For professional services firms, the challenge will be to reconcile their desire to be seen as innovators with the fundamental requirement that their work remains grounded in reality. Until such firms can demonstrate that they have mastered the technology they are so eager to sell, their reports will remain suspect—a dangerous position for any firm built on the foundation of professional excellence.

In the final analysis, the most important lesson from this debacle is that while AI can simulate expertise, it cannot replicate the integrity, accountability, and critical judgment that define the human professional. For KPMG, the path to recovery involves not just retracting a report, but rebuilding the culture of oversight that ensures such a mistake can never happen again.