Web of Agents – A Tail and Tame of Controlled Autonomy for Physical Devices

What happens when autonomous AI agents leave the screen and start interacting with the physical world? Each agent may be designed to perform a simple task, but when agents communicate, collaborate, and make decisions autonomously, unexpected behaviours can emerge - from unsafe physical actions and privacy breaches to denial of service, manipulation, or agents simply refusing to respond. This is the Web of Agents: a world where intelligent agents operate not in isolation, but as an interconnected ecosystem of physical and digital systems. In this interactive workshop, a team from Queen's University Belfast, Dr Vishal Sharma and Dr Gyan Wickremasinghe, together with Fannya R. Sandjaja, Priyal Gupta, Su Thinzar Thaw, Matteo Melis, and Mark Megarry, will explore the challenges of controlled autonomy for physical devices. Through live demonstrations involving everyday robots, computers, sensors, and connected devices, participants will discover what can happen when autonomous agents interact, and what it takes to keep them safe, trustworthy, and under human control. We will explore: • Web of Agents - how autonomous agents communicate, cooperate, compete, and make decisions across physical and digital environments. • Physical-Layer Security - how the physical communication environment itself can become an attack surface, and how sensing, wireless communication, device identity, and signal-level protections can contribute to secure autonomy. • Controlled Autonomy -designing systems that can act independently while remaining bounded by clearly defined safety, security, and ethical constraints. • Human-in-the-Loop - understanding when humans must intervene, approve, override, or take control of autonomous agents. Rather than simply watching AI make decisions, participants will become part of the system - taking the role of the human controller, security monitor, and safety engineer.