A partner at the Big Four accounting firm KPMG has received a $10,000 (£7,200) penalty for utilising artificial intelligence (AI) to deceive in an internal training course aimed at evaluating knowledge of the technology.
KPMG Australia mandated the unnamed audit partner to retake the exam after they uploaded training materials onto an AI platform to help answer AI-related inquiries. Andrew Yates, CEO of KPMG Australia, remarked in the Australian Financial Review: “Given the common use of these tools, some individuals violate our policy. We take these breaches seriously. We are also examining ways to enhance our approach within the existing self-reporting framework.”
The training materials instructed staff to download a reference manual for the course, but this partner breached the firm’s policies by uploading the reference document to an AI tool to respond to an exam question. This situation arises as KPMG improves its systems to identify AI-assisted cheating after a notable pattern of cheating on internal tests from 2016 to 2020.
Exam misconduct is not limited to Australia. The US regulator assessed fines totaling $8.5 million (£6.2 million) to the Dutch branches of Big Four firms Deloitte, PwC, and EY in June of the previous year due to “widespread exam misconduct.”
In 2022, the US watchdog imposed a fine of $7.7 million (£6.3 million) on KPMG UK and penalised four auditors for not addressing a major cheating scandal in the UK.
Yates continued, “Like many organisations, we have been contending with the role and utilisation of AI concerning internal training and assessment. It’s quite challenging to manage, given the rapid adoption of it by society.”
Deloitte Australia, another competitor within the Big Four, also experienced difficulties with AI. The firm provided a partial refund to the Australian federal government after submitting a report that contained errors linked to AI. A spokesperson from the government department stated, “Deloitte conducted the independent assurance review and has acknowledged that some footnotes and references were incorrect.”
You must be logged in to post a comment.