Google's gemini AI is a tool that can create audio with video, prepare picture books for children and even plan your travel. But along with its bright side, it also has a dangerous side which a research team brought out in a small but shocking demo.
From science fiction to reality
You must have seen in films that someone sitting far away controls the lights of the house or changes the temperature. But the researchers actually did this feat and that too by hacking gemini AI. They showed how by connecting AI to a smart device, not only changes can be made in the house but serious damage can also be caused by it.
According to WIRED's report, the team hacked the gemini Assistant running on an Android phone through a google Calendar invite in which dangerous code was hidden. As soon as they asked gemini to summarize the calendar details, the code got activated and started affecting all the devices in the house. In this way, a simple looking calendar invite became a means to take over the entire smart system of the house.
The scary limits of AI
Although the researchers did not reveal the technical details, this demo was enough to show how destructive AI can prove to be in the wrong hands. Smart devices have already been victims of cyber attacks and now with the advent of LLMs (Large language Models), the threat has increased manifold. This is why it is very important to ensure the security of these models before expanding smart home AI.
Google's response and future challenges
This hacking demo was shared with google and the company believes that controlling AI through home devices can be dangerous. Although google says that this vulnerability has not been misused so far, it is working on removing it. There are plans to bring gemini to smartwatches, smart TVs and even cars in the future, which was announced in the keynote of I/O 2025. Hopefully google and its security team will fix these vulnerabilities before cyber criminals take advantage of it.
This hacking demo was shared with google and the company believes that controlling AI through home devices can be dangerous. Although google says that this vulnerability has not been misused so far, it is working on removing it. There are plans to bring gemini to smartwatches, smart TVs and even cars in the future, which was announced in the keynote of I/O 2025. Hopefully google and its security team will fix these vulnerabilities before cyber criminals take advantage of it.
click and follow Indiaherald WhatsApp channel