Gemini AI flaws could have exposed your data

Gemini AI flaws could have exposed your data, with security researchers uncovering three vulnerabilities in Google’s Gemini AI assistant, according to Tenable. The flaws affect three Gemini components: Gemini Cloud Assist, Gemini Search Personalization Model, and Gemini Browsing Tool.

Hidden prompts inside web requests could be used to sneak malicious instructions into the system, potentially giving attackers control over cloud resources; the personalization model could inject prompts into a user’s Chrome browsing history to force the AI to leak personal data, and the browsing tool could transmit stored information to a malicious server via its page-summarisation feature.

Google fixed these issues by blocking Gemini from rendering dangerous links and strengthening defenses against prompt injections. If you relied on Google services that use Gemini, there is a chance the vulnerabilities were exploited before the patch, especially after visiting a malicious site or using Gemini features linked to cloud services. Posted: 1 October 2025.