Story perspectives

Google's Gemini AI Faces Major Security Flaws

10/3/2025

46 5

Summary

Story summary

Google’s Gemini artificial intelligence security issue, called 'Trifecta,' includes three vulnerabilities enabling data manipulation and unauthorized access.
Flaws affect Gemini Cloud Assist, Search Personalization Model, and Browsing Tool.
Attackers could inject commands and extract data; Google patched the issues, though risk existed before.
Experts advise treating AI as an active target and monitoring for unusual outbound requests to deter prompt-injection.

1 of 1

Story summary

Google’s Gemini artificial intelligence security issue, called 'Trifecta,' includes three vulnerabilities enabling data manipulation and unauthorized access.
Flaws affect Gemini Cloud Assist, Search Personalization Model, and Browsing Tool.
Attackers could inject commands and extract data; Google patched the issues, though risk existed before.
Experts advise treating AI as an active target and monitoring for unusual outbound requests to deter prompt-injection.