1/2/2026AI Engineering

Unitree G1's Critical Security Flaws: Remote Code Execution and Unauthorized Telemetry Exposed

Unitree G1's Critical Security Flaws: Remote Code Execution and Unauthorized Telemetry Exposed

Security researchers have discovered devastating vulnerabilities in Unitree’s robotic fleet, including remote code execution via Bluetooth and unauthorized telemetry transmission – raising serious concerns about the security architecture of one of robotics’ most popular R&D platforms.

The Critical Vulnerabilities

Two major security flaws have been identified in Unitree’s entire robot lineup, including the popular G1 model:

    • Remote code execution via Bluetooth Low Energy (BLE) using hardcoded AES keys
    • Unauthorized telemetry data transmission averaging 1.4MB per minute

The Bluetooth Attack Vector

The most severe vulnerability stems from Unitree’s implementation of Bluetooth Low Energy. The main control board runs BLE constantly with no user control over this functionality. More critically, all Unitree robots use identical hardcoded AES keys – a fundamental security anti-pattern that violates basic principles of secure system design.
Attackers within Bluetooth range can:

  • Connect to any Unitree robot using the shared AES keys
  • Inject arbitrary commands via the WiFi credential update mechanism
  • Execute commands with root privileges on the main control board

The Telemetry Problem

Despite Unitree’s claims that their robots are “designed for offline use”, testing reveals consistent unauthorized data transmission, similar to issues previously identified in other autonomous systems.

Claim Reality
Offline by default Automatically connects and transmits when WiFi credentials exist
User authorization required No explicit authorization mechanism found
Only basic telemetry Continuous bi-directional data transfer at 1.4MB/minute

Technical Impact Analysis

The severity of these vulnerabilities cannot be overstated. The combination of remote code execution and unauthorized telemetry creates a perfect storm of security risks that breaks fundamental safety assumptions about robotic systems.
For academic and research institutions using Unitree platforms, this presents immediate challenges:

  • Anyone within Bluetooth range can compromise demos and experiments
  • No way to verify what data is being collected and transmitted
  • Root access could enable malicious code persistence

Unitree’s Response

Unitree’s official statement acknowledges “security vulnerabilities and network-related issues” but fails to address the remote code execution vulnerability. Their focus on defending telemetry collection while ignoring the more critical RCE issue suggests a concerning disconnect from fundamental security principles.

Mitigation Strategies

For existing Unitree owners, options are limited:

  • Physically disable Bluetooth (requiring hardware modification)
  • Keep batteries removed when not in use
  • Disable WiFi connectivity (limiting functionality)
  • Wait for official patches (timeline uncertain)

Looking Forward

This security disaster highlights the risks of closed-source robotics platforms with proprietary control systems. As the robotics industry matures, transparency and security-first design principles must become standard practice – not optional features bolted on after vulnerabilities are exposed.