TL;DR
A KPMG Australia partner has been fined A$10,000 ($7,000) for using AI tools to cheat on an internal training course about AI. More than two dozen staff have been caught doing the same this financial year, highlighting a growing problem across professional services firms.
The Irony of AI Cheating on AI Tests
A partner at KPMG Australia has been forced to redo an internal training course and pay a A$10,000 fine after uploading course materials into an AI platform to generate answers. The training course itself was about the responsible use of AI — making the breach particularly embarrassing for the Big Four accounting firm.
The unnamed partner is not alone. KPMG says more than two dozen staff members have been caught using AI tools to pass internal exams during the current financial year. The firm has introduced detection measures and plans to publish figures on AI misuse when it releases annual results.
A Sector-Wide Problem
The incident fits into a broader pattern across professional services. The Association of Chartered Certified Accountants, the world’s largest accounting body, scrapped remote testing late last year because its safeguards could not keep pace with cheating technology. All four of the Big Four accounting firms have faced fines over cheating scandals across multiple countries in recent years.
“Like most organisations, we have been grappling with the role and use of AI as it relates to internal training and testing,” said Andrew Yates, KPMG Australia’s chief executive. “It’s a very hard thing to get on top of given how quickly society has embraced it.”
The matter drew political attention during an Australian Senate inquiry into industry governance. Greens senator Barbara Pocock called the situation “extremely disappointing” and criticised what she described as a “toothless system.”
The Australian Securities and Investments Commission confirmed the incident but said it would not take further action until the relevant professional trade body begins disciplinary proceedings. Under current rules, audit firms are not obliged to report such misconduct — the onus falls on individual partners to self-report.
Looking Forward
The case raises uncomfortable questions for organisations rolling out AI training programmes. If the tools being taught about are also the tools being used to cheat, companies may need to fundamentally rethink how they assess AI competency. Expect more firms to move toward supervised, in-person testing as remote assessment becomes harder to secure against increasingly capable AI tools.