Researchers have discovered that hackers can embed inaudible audio commands within seemingly normal audio files, such as podcasts or videos. These hidden commands can then be used to hijack AI voice chatbots, causing them to perform unintended actions. This vulnerability highlights a new potential attack vector for AI systems that rely on voice input. AI
Summary written by gemini-2.5-flash-lite from 1 sources. How we write summaries →
IMPACT Exploitable vulnerability in voice AI could lead to unauthorized actions and data breaches.
RANK_REASON The cluster describes a newly discovered vulnerability in AI voice chatbots, detailing a method for exploitation. [lever_c_demoted from research: ic=1 ai=1.0]