Conan Dewitt

Papers

Total Citations

H-Index

About

Conan DeWitt is a rising researcher at the forefront of embodied AI security, specializing in the vulnerabilities of large language model (LLM)-integrated robotic systems. His most-cited work, "A Study on Prompt Injection Attack Against LLM-Integrated Mobile Robotic Systems" (2024, 5 citations), pioneers the critical examination of how adversarial prompts can compromise robots that rely on models like GPT-4o for multi-modal decision-making. By demonstrating that seemingly benign inputs can hijack a robot’s context-aware responses, DeWitt has exposed a fundamental security gap in the fusion of LLMs with physical agents—a contribution that is already informing safer system designs. His research bridges cybersecurity, robotics, and natural language processing, offering both a wake-up call and a roadmap for developing robust, attack-resistant architectures. Though early in his career, DeWitt’s work has been recognized for its timely relevance, earning him invitations to speak at workshops on trustworthy AI. For students and researchers, his studies serve as a compelling entry point into the emerging field of adversarial machine learning in robotics, highlighting the urgent need to secure the next generation of autonomous systems.