When Healthcare IoT Applications Meet Agentic AI: Emerging Security and Privacy Challenges

In recent years, Healthcare Internet of Things (HC-IoT) applications have enabled remote patient monitoring, real-time diagnostics, and personalized treatment. A new shift is now underway with the emergence of agentic AI, which is moving HC-IoT from passive analytics to goal-driven autonomy, where AI agents will not only interpret data but also select tools, coordinate with other services and agents, and trigger actions to perform different tasks. Despite the potential benefits of this technology, it also expands the attack surface beyond classic sensing-and-communication threats by creating emerging security challenges such as multi-step tool misuse, cross-layer cascading failures, unsafe action chains, and coordination risks across multiple agents. Given that, we want to present a comprehensive survey paper regarding the HC-IoT applications’ security challenges in the context of agentic AI. We explain how autonomy changes the attack surface in different HC-IoT applications. A single compromised device can steer a closed, step-by-step decision loop, turning attacks from simple data corruption into real action-level harm. Next, we talk about how the risks extend beyond models and networks to the agent’s memory, planning logic, tool interfaces, and interactions. We also provide a clearer architectural view by distinguishing reactive, deliberative, and hybrid agent pipelines, and linking each to its main failure modes and attacker entry points. In the next phase, we review existing defenses and explain why standard security measures are incapable of handling the security challenges of agentic AI-enabled HC-IoT applications at the decision and action layers, as well as the communication layer. We also highlighted open research challenges alongside possible future directions, which could ensure the future of this technology. Finally, we want to acknowledge that this work provides a roadmap for researchers, students, industry stakeholders and policymakers how they can secure the next generation of intelligent, autonomous healthcare systems.