A prompt injection flaw in Google Gemini's voice assistant let attackers hide malicious commands in notifications, enabling social engineering and more.
Malicious Notifications Could Trick Google Gemini Users
Summary + source link. No paywalls, no tracking.