Google’s Gemini for Workspace Vulnerable to Prompt Injection Attacks

Hidden Layer researchers demonstrated that Google’s Gemini for Workspace is vulnerable to indirect prompt injection across Gmail, Slides, and Drive, allowing attackers to craft emails, speaker notes, or shared documents that manipulate AI-generated responses and enable phishing or misinformation; Google currently labels these behaviors as "intended," raising concerns about integrity and trustworthiness of LLM-driven features.