Announcements from the Robotic Hacking Community — DEF CON recaps, event news, and coverage from the hacker community ecosystem.
A robot can function as expected relative to the information it receives and still produce an unsafe outcome if that information has been manipulated. Cyber threats add an adversary to the safety equation. They can turn an unlikely event into a repeatable one.
Building on findings from the Physical AI Safety Stress Test CTF at the Robotic Hacking Village at DEF CON 34, VicOne released a free extension for NVIDIA Isaac Sim. Developers can load their own robot model and intended task, apply curated cyber-safety attack scenarios, check whether a vision-language model holds up against adversarial inputs, and repeat the tests as their systems change.
At the event, 16 teams recorded 25 successful solves across challenges covering AI models, perception, cloud and API services, robotics communications, and embedded systems. Nearly every core AI and robotics challenge was solved at least once, while one advanced embedded systems challenge involving cryptography and firmware went unsolved.
The extension is a free download for supported versions of NVIDIA Isaac Sim. It sits alongside VicOne's paid Radeis platform, which covers broader pre-deployment validation.
VicOne announced it would bring its Physical AI Safety Stress Test to DEF CON 34. Teams work through hands-on missions in six areas: prompt injection against robot AI reasoning, poisoned AI policy models, adversarial visual inputs, cloud and API weaknesses, ROS 2 and DDS communication weaknesses, and embedded firmware trust-chain weaknesses.
The challenge runs in a Real2Sim digital twin built in a physics-based robotics simulator, modeling the Reachy Mini Lite and the LeKiwi mobile manipulator. Each team gets an isolated environment through a challenge API, so no live robots are exposed. The CTF was held at the Las Vegas Convention Center, West Hall 4, Level 1, Room 1309, August 6–9, 2026, and was limited to 80 teams.