
In recent years, the rise of generative AI technology has transformed the landscape of the tech industry, making it nearly impossible to avoid its influence. While major companies like Google prioritize discussions on AI safety, the rapid development of AI capabilities has also introduced new malware threats, which researchers from Tel Aviv University have dubbed "promptware." In a groundbreaking study, the researchers revealed how they successfully manipulated Google’s Gemini AI to control smart home devices using seemingly harmless calendar appointments. This incident marks a significant moment as it represents one of the first known instances of an AI-related attack having tangible effects in the real world. Gemini's integration within the Google ecosystem provides it with a variety of functionalities, including calendar access, interaction with Assistant smart home devices, and the ability to send messages. This interconnectedness makes it an attractive target for cybercriminals who aim to disrupt systems or compromise sensitive information. The researchers employed an indirect prompt injection attack, a method where malicious instructions are fed into the AI by an external party rather than the user. The process begins with a calendar event that contains concealed harmful instructions. When a user requests a summary of their schedule, Gemini inadvertently processes this tainted event, leading to unexpected actions. An example of this exploit involved instructing Gemini to respond to specific phrases from the user to control devices like lights and thermostats. This innovative attack cleverly circumvented Google’s current security measures by linking harmful commands to benign user interactions with the AI. The findings from this research suggest that controlling any device connected to Google could be achieved through this method, highlighting a new vulnerability in smart home networks. The team anticipates that this incident may represent a pivotal shift, transitioning prompt-injection attacks from the digital realm into the physical world.
As the accessibility of powerful large language models (LLMs) increases, the internet faces an escalating threat from sp...
TechCrunch | Jul 06, 2026, 16:05
Microsoft has announced significant layoffs, cutting approximately 4,800 positions, which amounts to 2.1% of its global ...
TechCrunch | Jul 06, 2026, 15:25
In a rare disclosure, Canada's Communications Security Establishment (CSE) has revealed its engagement in a series of st...
TechCrunch | Jul 06, 2026, 15:05
In a significant restructuring move, Microsoft has announced the elimination of 4,800 jobs, which accounts for 2.1% of i...
CNBC | Jul 06, 2026, 13:50
In a significant move, Apple has reinstated card payment options for Apple Account purchases in India, marking a pivotal...
TechCrunch | Jul 06, 2026, 16:40