Hackers have uncovered a critical security flaw in Google Gemini AI, through which even a seemingly harmless calendar entry could be enough to take control of smart home systems. The vulnerability was disclosed in August 2025, and it became clear that attackers could bypass standard authentication processes and gain access to users’ devices and data.
The weakness was demonstrated by researchers from Adversa AI at the Black Hat 2025 conference, highlighting how attackers could exploit Gemini for Workspace to process and execute malicious instructions embedded in calendar events. This allowed hackers to seize control of smart home devices – including thermostats, cameras and smart locks – as well as to access personal documents. The risk was compounded by the fact that the Gemini API integrates with more than seventy Google services, meaning the attack could be carried out on a broad scale without requiring administrator privileges or advanced technical skills.
Google responded in mid-August by swiftly issuing a patch, stressing that there was no evidence of the flaw being exploited in real-world conditions. However, Alex Polyakov, founder of Adversa AI, warned that the case illustrates the wider security risks of AI assistants when granted access to users’ devices and data. He emphasised that strengthening defence mechanisms and improving user awareness are vital to protecting against similar attacks in the future.
Sources:
Matt Burgess
Vlad CONSTANTINESCU
