The Unitree G1 EDU humanoid robot, a star in research labs, can be compromised and remotely activated via two critical vulnerabilities—one of which doesn’t even require being on the same network, just being within Bluetooth range.

Hacking a Humanoid Robot via Bluetooth: Unitree G1 EDU Flaws Open a New Era of Physical Security

What's Happening

The Unitree G1 EDU is one of the most accessible humanoid robots on the market—a 35 kg precision machine with two legs, two arms, sensors everywhere, sold as an open research platform. It can be found in dozens of university labs and R&D centers worldwide.

Two flaws have just been revealed in its architecture. The first: a vulnerability allowing root access via RCE—full control over the operating system, actuators, and sensors. The second: Bluetooth activation without authentication—anyone within Bluetooth range can trigger system actions. No password, no pairing, nothing.

Why These Flaws Are Unique

We see RCE vulnerabilities every week. But unauthenticated Bluetooth activation on a 35 kg physical robot? That’s different: the risk is no longer just digital—it becomes physical.

A compromised humanoid robot in a lab could be remotely commanded to move, grab objects, or simply fall—with real-world physical consequences. This isn’t science fiction: it’s the direct, immediate risk of these vulnerabilities in real-world use.

High-risk scenarios include university labs with semi-open access, public demos at tech expos, and hackerspaces or fab labs where physical access is hard to control.

Robotics Security: A Massive and Urgent Challenge

These two flaws highlight a structural lag: robotics security is where computer security was 20 years ago. Manufacturers optimize for performance, ease of development, and stability. Security comes last.

The Bluetooth flaw is particularly telling: the mobile industry solved Bluetooth authentication issues in the 2000s–2010s. Seeing them reappear in 2026 in robotics shows that every new tech domain must relearn security lessons the hard way.

As humanoid robots move out of labs and into warehouses, hospitals, and public spaces, these flaws will stop being academic. The era of robotics security starts now—late, as always.

What to Do Now

Immediate Actions

1. Check Unitree security bulletins for a firmware patch—apply it immediately if available.
2. Disable Bluetooth if unused: on embedded Linux, rfkill block bluetooth—evaluate based on experimental needs.
3. Control physical access: until patched, the robot must be in a strictly controlled space.
4. Isolate the robot on a dedicated network: no connection to sensitive lab resources from the G1 EDU.
5. Report to your security team if your institution has G1 EDU units—this is a documented physical risk, not a hypothetical one.