Black Hat Demo: Simple QR Code Breaks Gemini and Hijacks Unitree Robot
MarcoFigueroa · x · 2026-08-09
At the Black Hat USA security conference, researchers demonstrated an attack technique dubbed Kinetic Prompt Injection.
By using a simple QR code, the researchers successfully executed a prompt injection attack against Google's Gemini model. Due to vulnerabilities in how LLMs process multimodal inputs, the malicious instructions embedded in the image were parsed, causing a Unitree robot powered by the model to exhibit erratic behavior and act like a "rabid attack dog."
This research highlights that as AI agents are integrated into the physical world, prompt injection attacks can transcend digital data exfiltration, resulting in actual physical harm with a real-world blast radius.
More from Embodied
- VLA-Δ: Zhejiang University Adds Physical State Transition Supervision to VLA Models — OmniAI-ZJU · 2026-10-16
- PINT: Shared Predictive Intent Tokens Unify Reasoning, World Modeling, and Robot Actions — Xinyu Guo · 2026-10-16
- TouchScale releases 500 hours of unified egocentric vision-tactile data for scaling study — jiqizhixin · 2026-10-16
- Ghost raises $11M for Core, a personal AI computer that runs locally in your home — nick_linck · 2026-10-16
- Robot shows up at Costco to fetch the famous $1.50 hotdog — majidmanzarpour · 2026-10-16
- 15 Leju Kuavo humanoid robots run 2 months at Hangzhou auto-parts line, 98% success rate — CyberRobooo · 2026-10-16