AI Can Steal Passwords by Analyzing VR Avatar Hand Motions
chrisgrayson · x · 2026-08-12
AI can infer passwords and sensitive information typed on physical keyboards by analyzing the subtle hand movements of avatars in Virtual Reality (VR) meetings.
Researchers at the University of Chicago demonstrated these keyboard inference attacks in Meta Horizon Workrooms. Because VR headset sensors cannot perfectly track finger motions, physical typing translates into subtle avatar hand movements, which AI can capture and analyze to reconstruct keystrokes.
The simplest method requires no data access from the target; the attacker only needs to enter the same virtual room and visually record the target avatar's hand motions.
More from Safety
- Grok Bot Integration Faces Hurdles: Call for Verified AI Agents on X — Daniel_Farinax · 2026-08-12
- AI-Powered Scams Are Getting Highly Personalized and Hard to Detect — SpencrGreenberg · 2026-08-12
- Research: Encrypted CoT Traces Can Be Extracted via Weaker Sibling Models — tw1st3d_m3nt4t · 2026-08-12
- The hidden costs of an AI pause and the limits of AI labs' competence — sebkrier · 2026-08-12
- Eyro Introduces Zero-Trust Agent Architecture Immune to Prompt Injection — IntelligentFigure441 · 2026-08-12
- AI Capabilities Overstated? Researcher Debunks 6-Sigma Knowledge Worker Prediction — QuintinPope5 · 2026-08-12