- An audit partner was fined £7,200 for breaching AI policies.
- The test was designed to assess staff understanding of artificial intelligence.
- Firms are tightening controls as AI misuse becomes harder to detect.
A partner at KPMG has been fined £7,200 ($10,000) after using artificial intelligence to help complete an internal training test meant to assess knowledge of the technology itself. The case sheds light on the growing challenge firms face as AI tools become part of everyday work.
KPMG Australia required the unnamed audit partner to retake the test after discovering they had uploaded training materials to an AI platform to generate answers to questions. The course had advised staff to download a reference manual, but the firm’s policy prohibited uploading such documents into external AI tools.
Andrew Yates, chief executive of KPMG Australia, reportedly said that given the widespread use of these tools, some people breach company policy and the firm takes such cases seriously, as quoted in a news report. He added that KPMG is looking at ways to strengthen its approach under the current self-reporting system.
A wider problem across the industry
The incident comes as firms step up efforts to detect AI-related cheating, following earlier cases of exam misconduct between 2016 and 2020. Concerns around test integrity have not been limited to Australia. In June, a US regulator fined the Dutch arms of Deloitte, PwC and EY a combined $8.5m (£6.2m) over what it described as widespread exam misconduct.
In 2022, regulators in the US also fined KPMG UK $7.7m (£6.3m) and sanctioned four auditors over several failings, including not properly addressing a broader cheating scandal.
Yates reportedly said many organisations are still trying to understand how AI should be used in training and testing, noting that the rapid pace of adoption has made it difficult to manage effectively, as quoted in a news report.The issue extends beyond exam settings. Deloitte Australia previously issued a partial refund to the federal government after submitting a report that contained errors linked to its use of AI. A government spokesperson reportedly said the firm confirmed that some footnotes and references were incorrect, as quoted in a news report.
The developments suggest professional services firms are still working out how to balance the benefits of AI with the risks it brings, particularly when accuracy and trust sit at the core of their work.
