Understanding the Ethical Risks of AI: How Delegation Can Encourage Dishonest Behavior

In a recent Navex blog post, Matt Kelly, CEO of Radical Compliance, examines new research in Nature that reveals one of the most overlooked ethical risks of AI in the workplace: how delegating tasks to AI can encourage unethical behavior. While compliance professionals often focus on bias, privacy, and accountability, Kelly underscores that the use of AI itself can change how people behave, sometimes for the worse.

The Nature study, involving 8,000 participants across 13 experiments, found that people were more likely to lie when interacting with AI than when reporting results directly. For instance, when participants rolled dice for money, 95% were honest without AI, but only 75% were honest when using AI to report outcomes. When given freedom to direct the AI’s goals, 84% instructed it to prioritize personal gain. The researchers concluded that AI creates “moral distance,” allowing people to rationalize unethical decisions they might not otherwise make.

Kelly warns that this distancing effect presents a serious governance issue. Compliance officers should integrate the ethical risks of AI into corporate policies, training, and risk assessments. According to the NAVEX 2025 State of Risk and Compliance Report, most compliance leaders are already involved in AI governance, positioning them to address these behavioral risks directly.

For risk managers, AI oversight isn’t just about algorithms but about human behavior. Embedding ethical risk considerations into AI governance charters and controls can help organizations mitigate these emerging risks before they erode integrity and trust.