Skip to main content
COSMICBYTEZLABS
NewsSecurityHOWTOsToolsTraining
StudyProjectsNewsletterHire MeAbout
Subscribe

Press Enter to search or Esc to close

News
Security
HOWTOs
Tools
Training
Study
Projects
Newsletter
Hire Me
About
RSS Feed
Reading List
Subscribe

Stay in the Loop

Get the latest security alerts, tutorials, and tech insights delivered to your inbox.

Subscribe NowFree forever. No spam.
COSMICBYTEZLABS

Your trusted source for IT intelligence, cybersecurity insights, and hands-on technical guides.

2567+ Articles
161+ Guides

CONTENT

  • Latest News
  • Security Alerts
  • HOWTOs
  • Checklists
  • Projects
  • Exam Prep

RESOURCES

  • Search
  • Browse Tags
  • Newsletter Archive
  • Reading List
  • RSS Feed

COMPANY

  • About Us
  • Contact
  • Privacy Policy
  • Terms of Service

© 2026 CosmicBytez Labs. All rights reserved.

System Status: Operational
  1. Home
  2. News
  3. Two Unitree G1 EDU Humanoid Robot Flaws Enable Root RCE — One Starts Over Bluetooth
Two Unitree G1 EDU Humanoid Robot Flaws Enable Root RCE — One Starts Over Bluetooth
NEWS

Two Unitree G1 EDU Humanoid Robot Flaws Enable Root RCE — One Starts Over Bluetooth

Researcher discloses two root RCE chains in the Unitree G1 EDU robot; one is a wormable, unauthenticated Bluetooth exploit reaching root remotely.

Dylan H.

News Desk

August 28, 2026
4 min read

What Happened

Security researcher Olivier Laflamme (known as Boschko) has disclosed two independent root remote code execution chains affecting the Unitree G1 EDU humanoid robot — including a Bluetooth Low Energy (BLE) path that lets an unauthenticated attacker reach root on the robot's Locomotion PC from within radio range, no pairing required. The Locomotion PC runs a real-time Linux kernel with root privileges over motors, cameras, audio, and voice.


Chain #1: CVE-2026-76639 — Network-Adjacent RCE via chat_go/bashrunner

AttributeValue
CVSS Score8.8 (High)
Attack VectorNetwork-adjacent
ImpactRoot command execution

This chain combines three weaknesses:

  1. An unauthenticated WebRTC-to-DDS bridge on TCP port 9991 provides the initial access point.
  2. A static AES-128 key, world-readable on the device, facilitates interaction with internal services.
  3. A path traversal flaw in the chat_go knowledge-upload API lets an attacker plant a file into the bashrunner service's whitelist directory.

Because bashrunner builds its execution whitelist at import time and runs files by shell rules without checking extensions, an attacker can publish DDS control messages over the unauthenticated bridge to restart bashrunner, then execute the planted payload — as root.


Chain #2: CVE-2026-76640 — "UniBLEed" Bluetooth Root RCE

This is the more severe chain: it takes an attacker from Bluetooth range to root on a $20,000 humanoid robot, crossing BLE, Unitree's cloud API, the companion mobile app, and the robot's own firmware.

How it works:

  • The BLE attack starts at GATT characteristic 0xFFE2, which accepts writes with only basic WRITE permissions — no BLE pairing needed to talk to the robot.
  • A plaintext bootstrap request returns a robot-specific AES-128 key wrapped in RSA encryption.
  • Unitree's /device/bindExtData cloud API can decrypt that RSA-wrapped material for any authenticated Unitree account — the endpoint reportedly never verified that the submitted serial number actually belonged to the requesting account.
  • With the recovered AES key, an attacker establishes the authenticated BLE state needed for Wi-Fi provisioning, then leverages a separate path traversal in the robot's AI chatbot knowledge base (which leaks the binary's load address) plus a 1,050-byte BSS buffer overflow to corrupt the event loop into calling system() as root.

Laflamme notes the exploit is wormable: once one G1 is compromised, it can propagate the same exploit to the next G1 within range, indefinitely.


Disclosure and Fix Status

  • Unitree patched the cloud account-to-robot ownership check in July 2026, closing the path where any authenticated account could pull another robot's key material.
  • Both chains were verified with Unitree during coordinated disclosure, and the key-retrieval flow now checks ownership before returning plaintext key data.
  • No confirmed firmware release fixes either CVE outright — Unitree has not published a specific patched firmware version in accessible guidance, so G1 EDU owners currently lack a firm update target.
  • The issues were reproduced across four physical G1 robots; Laflamme was awarded $6,700 in combined bug bounties across the two CVEs.
  • This research builds on a prior disclosure ("UniPwn") into Unitree BLE/RCE issues, with this new "UniBLEed" work representing an expanded, fully unauthenticated attack chain.

Why This Matters

Humanoid robots are increasingly deployed in research labs, universities, and industrial pilots, often on shared networks or in publicly accessible spaces. A wormable, unauthenticated root exploit reachable over Bluetooth — with no user interaction and no pairing — turns a single compromised unit into a foothold for propagating to every other G1 within radio range. Until Unitree ships a confirmed firmware fix, organizations running G1 EDU units should treat Bluetooth exposure as an active risk and restrict physical/RF proximity to untrusted parties where possible.


Sources

  • The Hacker News — Two Unitree G1 EDU Humanoid Robot Flaws Enable Root RCE, One Starts Over Bluetooth
  • Boschko — UniBLEed: Unauthenticated Root RCE on Any Unitree G1 Humanoid Robot Within Bluetooth Range
  • GBHackers — Unitree G1 Humanoid Robot Flaws Allow Unauthenticated Root RCE Over Bluetooth
#Unitree#Robotics#RCE#Bluetooth#IoT Security#CVE-2026-76639#CVE-2026-76640

Related Articles

As Global Powers Explore Humanoid Robots, Cyber-Risk Looms

Nation-states are racing to dominate the embodied AI and humanoid robotics market, but as governments and militaries integrate these systems, the…

5 min read

SentinelOne Device Control: Block USB & Bluetooth Devices

How to block USB drives and Bluetooth peripherals with SentinelOne Device Control — audit in Monitor mode, then enforce allow/block policies safely.

20 min read

Edu-Tech Firm Instructure Discloses Cyber Incident, Probes

Instructure, the company behind the widely used Canvas learning management system, has disclosed a cybersecurity incident and is investigating its scope....

4 min read
Back to all News