AI-rewritten: This is a summary of an article from The Conversation, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

Professor Murugan Anandarajan of Drexel University defines "latent risk" as the possibility that people lose capabilities organizations still depend on them to exercise. This occurs when AI performs reliably for months, leading employees to believe close scrutiny rarely changes outcomes. Consequently, employees may become less practiced at independent judgment and less prepared to intervene when the AI makes an error or cannot proceed.

This phenomenon builds on decades of research into automation bias, where people miss problems or follow automated recommendations despite contrary evidence. The current difference is that AI now assumes larger parts of analysis and decision-making, shifting work from doing tasks to overseeing AI actions. While companies like Deloitte are redesigning training to maintain necessary judgment, cognitive psychologist Lisanne Bainbridge noted decades ago that as systems take over work, people face new challenges in monitoring them rather than simply having lighter workloads.

Organizations often judge success by accuracy and productivity numbers, which may look good even as human roles change. For instance, fewer overrides might mean the AI is more accurate, not that employees are less ready to intervene. The core issue is that authority, responsibility, and capability can separate; an employee may remain responsible for outcomes they have less control over because they rarely exercise independent judgment themselves.

To address this, organizations must treat latent risk as a governance problem rather than just a human performance issue. Formal roles often do not match actual capabilities as AI improves. Anandarajan suggests organizations should periodically ask what relationship between people and AI they designed versus what they actually have today. They need to decide which human capabilities remain essential when AI reaches its limits, perhaps by occasionally asking employees to make decisions without seeing an AI recommendation first.