Elizabeth Waeni Mutisya

Fault Propagation and Risk Containment in Agentic AI Systems

Agentic artificial intelligence systems differ from conventional language model applications because they can reasonover a task, select tools, invoke external systems, observe resulting state, and continue acting toward a goal. Thisaction loop creates a security and reliability problem that cannot be adequately describ...