Root RCE Vulnerabilities Found in Unitree G1 EDU Humanoid Robots

Two critical root remote code execution (RCE) vulnerabilities have been discovered in the Unitree G1 EDU humanoid robot, potentially allowing unauthorized users to take complete control of the machine. The flaws, identified by security researcher Olivier Laflamme, expose the robot to attacks via both wireless and network-adjacent paths.

The vulnerabilities are tracked as CVE-2026-76639 and CVE-2026-76640. According to the research, CVE-2026-76640 enables root access through the robot's Bluetooth Low Energy (BLE) GATT server and its WiFi provisioning stack. Separately, CVE-2026-76639 provides a network-adjacent attack vector; this flaw leverages a path traversal vulnerability within a knowledge-base upload process, allowing an attacker to utilize the chat_go AI service and bashrunner to execute commands with root privileges.

The Research Context

The Unitree G1 EDU is specifically engineered for educational and research environments, where it is often deployed in labs or classrooms. Because these robots are designed for flexibility and integration, they rely on various communication interfaces to interact with developers and external systems. Laflamme's discovery highlights significant security gaps in how these interfaces—specifically BLE and internal network services—handle incoming data and authentication.

Implications for Robotics Security

Achieving root RCE on a humanoid robot is a severe security failure because it grants the attacker total authority over both the software and the physical hardware. In a practical scenario, this could allow a malicious actor to manipulate the robot's movements, creating immediate physical safety risks for anyone in the vicinity. Beyond physical danger, the Locomotion PC serves as a hub for sensitive research data, which could be exfiltrated once root access is established. Furthermore, a compromised robot can act as a persistent pivot point, allowing an attacker to bypass perimeter security and launch further attacks against other devices on the local network.

Next Steps

While the vulnerabilities have been disclosed, users of the G1 EDU should monitor for official firmware updates from Unitree to patch these gaps. The discovery underscores a broader trend in the robotics industry where the rapid deployment of AI-integrated hardware often outpaces the implementation of rigorous security hardening. Future audits will likely focus on whether similar path traversal or BLE flaws exist in other humanoid platforms used in academic settings.