Recent findings from the Black Hat computer-security conference highlight vulnerabilities in Gemini AI that could allow malicious prompts, known as “promptware,” to manipulate Google Home-connected smart devices. Researchers from Tel Aviv University demonstrated how subtle commands inserted into Google Calendar invites could instruct Gemini to control everyday devices like lights and windows. This represents a significant threat as Google integrates Gemini into its smart home ecosystem, transitioning from Google Assistant. Despite Google’s assurances of implementing fixes since receiving the report in February, concerns remain about potential exploits through email and calendar interfaces. Andy Wen from Google Workspace emphasized the importance of responsible disclosure in addressing these vulnerabilities. While smart home hacking is currently rare, the integration of generative AI into home systems raises new security questions. Users are advised to monitor their devices and consider disabling Gemini if concerned. To mitigate risks, further security enhancements are necessary as AI technologies evolve.
Source link
Share
Read more