KPMG Australia is dealing with 28 cases of staff using artificial intelligence to cheat on internal exams since July, including a senior partner who faces a financial penalty of more than $10,000 for breaching internal AI-use policies during mandatory training. The incidents emerge as the firm promotes itself as a leader in AI consulting and after years of prior cheating concerns on internal tests between 2016 and 2020, which prompted the business to overhaul how it designs, monitors and polices staff assessments.
The firm now actively tracks AI-related misconduct across more than 20,000 internal tests each year, covering a workforce of around 10,000 people with an annual turnover rate of about 20%. The flagged cases this financial year mostly involve employees at or below manager level but the partner case stands out because the person is a registered company auditor, a role subject to stricter rules due to its importance in safeguarding financial markets. The misconduct involved uploading training reference material into an AI tool to help answer an open-book exam question, an action allowed for private review but explicitly banned when used to generate solutions.
In response, KPMG has introduced monitoring systems designed to detect when staff rely on generative AI outside policy during tests and it has rolled out firm-wide education explaining where AI is acceptable and where it crosses the line into cheating. The partner’s conduct triggered an internal investigation, a fine linked to future income and self-reporting to a professional accounting body, while the wider group of staff involved face internal consequences under the firm’s conduct framework. KPMG has also voluntarily shared information with the corporate regulator as part of ongoing conversations, despite current rules not requiring firms to notify the regulator unless a professional body makes a disciplinary finding.
The situation highlights how professional services firms are struggling to set clear boundaries on AI use just as they earn a growing slice of their global revenue, about $US40 billion in KPMG’s case, from technology and AI-related services. By committing to separately disclose AI-related cheating in its annual results and by checking that staff actually self-report misconduct to their professional bodies, KPMG appears to be trying to set a new benchmark on transparency. However, these moves also increase pressure on other big firms and regulators to clarify reporting obligations, tighten oversight and decide how far internal AI monitoring should go as workplaces and education systems worldwide adapt to rapidly spreading AI tools.
