- Experts have warned that entry into the same calendar can quietly hijack your smart home without your knowledge
- Researchers proved that AI can be hacked to overcome smart homes by using only words
- Say “thank you”
The promise of AI-Integrated Homes has long been convenient, automation and performance, however, a new research by researchers at Tel Aviv has exposed another disturbing reality.
AI’s successful immediate injection may be an example of the real world of the attack, which the team used to use Google Calendar compromising in Gemini smart home.
The attack exploited the integration of Gemini with the entire Google Economic System, especially in it the ability to access calendar events, translate natural language indicators, and control smart devices associated.
You can like
From scheduling to sabotage: exploiting access to everyday AI
Gemini, though limited to independence, has enough “agent capabilities” to implement commands on smart home systems.
These contacts became a responsibility when researchers filed malicious instructions in the calendar appointment, it was masked as a regular event.
When the user later asked Gemin to summarize his schedule, he inadvertently mobilized the hidden instructions.
The embedded command included instructions to work as a Google Home Agent, unless the combined phrase like “thanks” or “definite” was typed by the user.
At that time, Gemini activated smart devices such as lights, shutters, and even a boiler, none of which had adopted the user at the time.
These delayed motivations were particularly effective in ignoring existing defense and confusing the source of actions.
This procedure, called “quickly”, raises serious concerns about how AI interface translates user input and external data.
Researchers argue that such urgent injection attacks represent the growing class of risks that combine social engineering with automation.
They showed that this technique could go far beyond the control control devices.
It can also be used to delete appointments, send spam, or open malicious websites, measures that can directly cause identification theft or malware infection.
The research team integrated with Google to show the risk, and in response, the company accelerated the rollout of new reservations against instant injection attacks, which includes additional examination of calendar incidents and additional confirmation of sensitive measures.
Nevertheless, questions remain about how these reforms are expanded, especially when Gemini and other AI systems gain more control over personal data and devices.
Unfortunately, traditional security suites and firewall protection are not designed for such an attack vector.
Safe USers, users should restrict access to AI tools and assistants like Gemini, especially calendars and smart home control.
Also, refrain from storing sensitive or complex instructions in calendar events, and do not allow AI to follow them without any supervision.
Be careful about extraordinary treatment of smart devices and disconnect access if something ends.
Through the wired
