Europe’s Big 4 can't stop making AI mistakes; PWC's report has ChatGPT as 'Source'

Europe’s Big 4 accounting and consulting firms have faced renewed scrutiny after a report by PwC revealed widespread use of AI-generated “hallucinations” in its Middle East thought leadership publications. Researchers from AI detection platform GPTZero, in collaboration with the Financial Times, identified multiple instances of fabricated footnotes, incorrect citations, and misleading references within four PwC reports published over the past two years. These reports addressed topics ranging from corporate strategies for “agentic” AI systems to government public service guidelines and market forecasts for electric vehicles in the Middle East. One report cited a non-existent study on air quality in Riyadh, while another referenced a nonexistent academic paper with specific authors. A third document incorrectly attributed a real-world success story about a JPMorgan AI initiative to a teenage blogger on Medium with just 280 followers, despite the initiative having occurred in 2017—years before ChatGPT was developed. Additionally, a single claim about traffic accidents was repeated across two pages, each time with different, erroneous citations. One footnote even included a URL with a tracking tag indicating the source material was retrieved via ChatGPT. In response, PwC Middle East stated it takes the accuracy of its published research seriously and is currently updating a limited number of supporting citations across the affected reports. However, the incidents underscore broader challenges facing the Big Four as they increasingly rely on generative AI tools to produce research and marketing materials. This trend has already led to similar controversies involving other members of the group. Earlier this year, KPMG retracted a global report on “Agentic Al” after major corporations and organizations named in the document disputed the authenticity of the claimed achievements. The report, titled Redefining Excellence in the Age of Agentic Al, featured numerous false claims and fabricated case studies. Similarly, EY Canada removed a report on loyalty reward programs from its website after researchers identified multiple inaccuracies, including AI-generated hallucinations and references to non-existent studies. The firm acknowledged reviewing the circumstances surrounding the report’s publication. Deloitte, too, has encountered issues related to AI-generated content. In November, the firm faced criticism for a major healthcare report prepared for the Canadian government, which assessed services in Newfoundland and Labrador. Multiple factual errors in the report were suspected to stem from the use of artificial intelligence tools. This marks the second time Deloitte has been implicated in such issues within a month. These incidents collectively suggest a growing concern about the reliability of AI-assisted research produced by leading professional services firms. While these firms continue to promote themselves as pioneers in AI integration, the recurring nature of these errors raises questions about the effectiveness of internal oversight mechanisms. The reliance on AI tools for generating content has become more pronounced as firms seek to maintain competitive advantage in rapidly evolving markets. However, the lack of transparency around how these tools are used, and whether human verification processes are adequately enforced, has left room for misrepresentation. In some cases, AI-generated content has inadvertently led to the inclusion of outdated or irrelevant information, undermining the credibility of the reports. The presence of URLs with tracking tags, such as utm_source=chatgpt.com, further complicates efforts to distinguish between genuine sources and AI-derived content. As the controversy unfolds, industry experts warn that the current approach to AI adoption in professional services may require stricter regulatory frameworks and greater accountability measures. The incidents involving PwC, KPMG, EY, and Deloitte highlight the urgent need for clearer guidelines on the ethical use of AI in producing authoritative research. Until these firms demonstrate stronger controls and more transparent practices, concerns about the integrity of their outputs will persist. The ongoing scrutiny serves as a cautionary tale for firms navigating the complexities of integrating AI into their operations.