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

Big Four accounting form PwC has come under fire for a reports that allegedly contains AI “hallucinations”.Big Four accounting and consulting giant PwC has come under fire after an investigation revealed that several of its middle eastern “thought leadership” reports contained AI-generated “hallucinations,” fabricated footnotes and chatbot citations, including a reference that listed ChatGPT directly in its source URL. The findings, uncovered by researchers at AI-detection platform GPTZero and verified by the Financial Times, highlight struggles among top consulting firms as they attempt to market AI advisory services while using generative tools to write their own research.The investigation flagged four PwC Middle East reports published over the past two years. The documents covered topics such as corporate strategy for “agentic” AI bots, government public service guides, and market forecasts for electric vehicles in the region.According to the report by the Financial Times, researchers found that a significant portion of the citations in the reports were flawed. For example, one report cited a study on air quality in Riyadh that appears to be completely hallucinated by AI as there is no trace of the paper exists in the referenced journal or under the listed authors.In a second incident, multiple footnotes linked to broken web pages or articles that contained no mention of the claims they were supposed to support. Thirdly, to support a “real-world success story” about a JPMorgan AI initiative, PwC cited a teenage blogger on Medium with just 280 followers. The actual initiative took place in 2017—five years before ChatGPT was launched.Moreover, a single claim regarding traffic accidents was repeated three times across two pages, with each instance featuring entirely different citations. In one glaring footnote, the URL for a cybersecurity report included the tracking tag utm_source=chatgpt.com, effectively revealing that the source material was surfaced directly through a ChatGPT prompt.In a statement to the publication, PwC Middle East said it “takes the accuracy of our published research seriously and is updating a limited number of supporting citations” across the flagged reports.The Big Four’s ‘AI struggles’ in reportsPwC is not the only member of the “Big Four” to face public embarrassment over unverified AI content. Previous investigations by GPTZero led rivals EY and KPMG to retract published research papers after similar hallucinations were uncovered. In June, KPMG pulled its global report on “Agentic Al” after major companies and organisations named in it complained that their supposed achievements were completely made up. The professional services giant published the report, titled "Redefining excellence in the age of agentic Al," with numerous false claims and bogus case studies, according to a report by The Financial Times.In May, Financial Times reported that EY Canada had to remove a report on loyalty rewards programmes from its website after researchers identified several mistakes. The report, which was allegedly Al-generated, apparently included Al hallucinations, inaccurate citations and references to studies that do not exist. The consulting firm said it was "reviewing the circumstances that led to this article's publication" following scrutiny over the report.In November, Deloitte landed in trouble for the second time in a month after it prepared a major healthcare report for the Canadian government. As reported by The Independent, the report, which was supposed to assess healthcare services in Newfoundland and Labrador, contained multiple factual errors that are believed to be generated by artificial intelligence tools.