Google has patched a vulnerability in Gemini AI.

Google recently patched three vulnerabilities in its Gemini AI assistant that could have enabled attackers to manipulate the system for data theft and other malicious actions. Researchers at Tenable demonstrated how indirect prompt injections could exploit Gemini’s Cloud Assist, Search Personalization, and Browsing Tool features. One method involved inserting malicious prompts into log files, which Gemini would later process and act upon, potentially exposing sensitive data.

Read more:

https://www.securityweek.com/google-patches-gemini-ai-hacks-involving-poisoned-logs-search-results/