KPMG has removed its report titled "Rethinking Excellence in the Age of Agent AI" from its websites following complaints from organizations about inaccurate or misleading statements regarding their use of the technology, according to FT.
The document was published in October 2025 and claimed that several major organizations were already using AI agents to automate complex business processes, including investment consulting, risk management, logistics, and healthcare.
UBS and Transport Companies Deny Claims
One prominent example was UBS. The report stated that the bank had integrated AI assistants into investment consulting, risk control, and compliance on a platform developed in collaboration with Microsoft.
A UBS representative told the publication that this information was "factually incorrect." Following the bank's inquiry, KPMG began removing the document from its resources.
Similar complaints arose from Swiss Federal Railways. The report claimed that the railway operator was using AI agents to plan and optimize trips based on passenger preferences and environmental factors. The company stated that this description was inaccurate.
Representatives from Transport for London also described KPMG's claims about using agent AI for managing transport congestion and coordinating services as misleading.
GPTZero Identifies Signs of AI Hallucinations
The inaccuracies were identified by the research group GPTZero, which specializes in analyzing AI-generated content. Experts noted that many of the disputed sections resembled classic hallucinations of language models—instances where the system generates convincingly presented but false information.
Particular attention was drawn to the case of the Greater Manchester NHS. KPMG claimed that the organization was using agent AI to predict readmissions and route patients. However, the source cited in the report only described the technology's application in a project for early lung cancer detection and made no such claims.
The Issue Affects the Entire Consulting Sector
KPMG stated that it takes the accuracy of published materials seriously and is conducting an internal investigation into the circumstances surrounding the errors. The company emphasized that its policy requires mandatory human review of materials and verification of independent sources.
This incident is another example of the challenges associated with the use of generative AI in professional services. In May, EY was forced to withdraw a similar report after discovering fictitious references and other errors. Earlier, the law firm Sullivan & Cromwell also reported inaccuracies caused by AI usage.
Risks for the Research Market
GPTZero CEO Edward Tian warned that errors in reports from major consulting firms could create a "secondary hallucination" effect, where false information begins to spread through media, analytical reviews, and studies from other organizations.
According to GPTZero, industry publications and major foreign media have already referenced the conclusions of the controversial KPMG report. Experts note that such cases undermine trust in the analytical materials of companies that simultaneously advise clients on AI implementation and managing associated risks.
In June, Anthropic CEO Dario Amodei urged U.S. authorities to tighten regulations on artificial intelligence.