KPMG partner fined for using AI to cheat on internal training
A partner at the Big Four firm KPMG has been fined $10,000 (£7,200) for using AI to cheat in an internal training course meant to test knowledge of the technology.
KPMG Australia has forced an unnamed audit partner to redo a test after they uploaded training materials to an AI platform to help answer AI-related questions.
Andrew Yates, chief executive of KPMG Australia, told the Australian Financial Review: “Given the everyday use of these tools, some people breach our policy. We take it seriously when they do. We are also looking at ways to strengthen our approach in the current self-reporting regime.”
The training materials recommended that staff download a reference manual for the course, but this partner violated the firm’s policies by uploading the reference document into an AI tool to answer an exam question.
This comes as KPMG upgraded its processes to detect AI cheating after widespread cheating on internal tests between 2016 and 2020.
Exam cheating was not limited to Australia, as the US regulator issued fines of $8.5m (£6.2m) against the Dutch arms of Big Four firms Deloitte, PwC, and EY after “widespread exam misconduct” last June.
Back in 2022, the US watchdog fined KPMG UK $7.7m (£6.3m) and sanctioned four of its auditors for a series of offences, including its failure to address a widespread UK cheating scandal.
Big Four facing AI troubles
Yates added, “Like most organisations, we have been grappling with the role and use of AI as it relates to internal training and testing. It’s a very hard thing to get on top of, given how quickly society has embraced it.”
Its fellow Big Four competitor, Deloitte Australia, faced issues with how it used its own AI, issuing a partial refund to the Australian federal government after it sent a report with several errors caused by AI.
Speaking at the time, a government department spokesperson said: “Deloitte conducted the independent assurance review and has confirmed some footnotes and references were incorrect.”