The Perils of Agency: How Developers Perceive, Prioritize, and Address Risks in Agentic AI Products

The study examines how 35 industry developers perceive and manage risks associated with agentic AI systems, which operate autonomously and adapt to complex environments. It highlights that developers prioritize product and business risks over societal implications, such as job displacement and privacy concerns, and often lack effective controls to mitigate agentic risks without limiting the very capabilities that make these systems valuable. This underscores a critical tension in agentic AI development between harnessing agentic functionalities and managing inherent risks.