The terrifying reality at GEEKCON: smart devices are being controlled by hackers.
The recently concluded Shanghai event, dubbed the “Security Geek Competition,” revealed the true hidden dangers of the future smart era.
When global white-hat hackers gather, they showcase far more than just “destruction”:
They provide a profound warning about the security boundaries of embodied intelligence (physical + AI).
Revealing: it could be “infected” and taken over.
At the scene, a seemingly sci-fi demonstration shocked the audience:
A contestant commanded a humanoid robot to say “Open Sesame,” and the robot immediately broke free from its original owner’s control, obeying the contestant’s instructions.
Researchers explained that this is not magic,
but rather exploiting a logical flaw in the robot’s internal model Agent, allowing arbitrary code execution through specific voice commands.
Even more concerning is that this control has a chain transmission capability—the compromised robot can “infect” other robots of the same model with commands.
The significance behind such demonstrations is particularly heavy
If robots are widely deployed as service, companionship, or production tools in the future.
Once maliciously taken over, the issue is no longer just an individual risk, but a systemic risk for the entire cluster.
As judge Tan Xiaosheng stated, the security of embodied intelligence must become the primary premise for industry proliferation.
Smart glasses were hijacked in just 70 seconds: privacy panic has arrived.
Equally noteworthy as robot security is the vulnerability challenge of smart glasses.
In a project titled “Your Glasses, My Eyes,” contestants completed the malicious code implantation on AI smart glasses in just 70 seconds.
After implantation, the device’s indicator light was turned off, and the user was completely unaware that the camera had been remotely controlled.
Everything they saw was being live-streamed to the attacker.
Even more seriously, some research teams pointed out that the microphones of certain smart glasses could also be silently activated, allowing attackers to listen to environmental sounds in real-time.
All of this happened without the user’s knowledge.
Drone remote control vulnerabilities exposed, low-altitude economic security risks are imminent.
Looking at drone security: at the conference, contestants conducted practical tests on four mainstream commercial drones and ultimately discovered vulnerabilities that could forge control signals.
Through forgery, attackers have the ability to remotely control the aircraft, even causing it to deviate from its flight path or crash.
Considering that drone logistics and air traffic businesses are rapidly expanding, these vulnerabilities, if abused, will pose a significant threat to the low-altitude economy.
Hackers control robots, glasses, drones: where will the next disaster strike?
Not only robots and drones, but mobile device security has also been compromised by hackers.
The research team spent nine months conducting a deep audit of the biometric authentication systems of several mainstream smartphones and found design flaws in some devices.
Attackers could even obtain the shared keys used for identity verification, thereby bypassing the biometric authentication mechanism.
During a live demonstration, a smartphone’s PIN code was extracted in just two minutes. This means that risks related to unlocking, payments, and personal data are far more vulnerable than imagined.
Can security ratings save the day? GEEKCON proposes a new standard for smart devices.
The organizers of GEEKCON 2025 did not shy away from these risks but instead proposed a forward-looking initiative:
To establish a quantitative rating system for smart device security. They hope to make “security” as clear and comparable as restaurant hygiene ratings.
This system has already begun preliminary implementation in areas such as smartphones, cameras, and smart locks.
The response to this conference has been strong both inside and outside the industry.Many companies are beginning to reassess the security boundaries of their product designs, and security researchers are realizing that future offense and defense will not only be at the software level.It is also a complex arena involving physical entities and AI models.
The GEEKCON organizing committee emphasized that they will continue to promote a responsible vulnerability disclosure mechanism.
Collaborating with manufacturers to fix issues and jointly safeguard the basic security bottom line of the smart era.
This conference has ended, but the alarm it rings for the world has not yet quieted. The deeper technology penetrates into life, the more obvious the convenience it brings, but the lurking security risks also increase.
The real challenge is not to make intelligence smarter, but to ensure it never deviates from the safety track we set.
If you have read this far, it means you have grasped the core of the article! If you found the content helpful, please give it a like and leave a comment and share it to support us! If you want to receive timely updates on AI tools and animation production, remember to star our public account⭐! Thank you for reading, and wish you a pleasant life!