Gemini vulnerability allows hackers to turn on lights, open curtains remotely
A serious vulnerability in Google Gemini has just been discovered, allowing attackers to control the victim's smart home with just a meeting invitation.
Báo Khoa học và Đời sống•08/08/2025
According to Wired, a team of cybersecurity researchers has found a way to exploit Gemini through the “indirect prompt injection” technique. The vulnerability started with a malicious meeting invitation sent to Google Calendar.
When a user asks Gemini to summarize their schedule, the AI chatbot accidentally executes a smart home control command. The Google Home system can be forced to open doors, turn off lights, or close curtains without the user knowing.
Google said it received the report in February and is working on a fix. Andy Wen, chief security officer at Google Workspace, warns that this type of attack will be around for a while. He also reassured that actual attacks remain extremely rare at the moment.
However, the rapid development of AI also makes the risk of exploitation increasingly difficult to control. Dear readers, please watch more videos : AI Trash Cleaning | Hanoi 18:00
Comment (0)