Generative AI systems have seen significant growth, introducing new malware threats like promptware. Researchers at Tel Aviv University exploited Gemini's capabilities by tricking it into manipulating Google smart home devices using a poisoned calendar appointment. By employing an indirect prompt injection attack, malicious instructions were hidden within calendar events. When users asked Gemini to summarize their schedule, the AI processed these harmful instructions and executed commands, demonstrating the potential risks associated with AI and its connectivity to various applications.
Researchers from Tel Aviv University demonstrated that they could trick Gemini into manipulating Google smart home devices through a series of malicious calendar appointments, marking a notable AI-related attack.
Gemini's integration with the broader Google app ecosystem makes it a vulnerable target for indirect prompt injection attacks that can lead to malicious and harmful outcomes.
Collection
[
|
...
]