Google Workspace’s continuous approach to mitigating indirect prompt injections

### Executive summary: This blog post explains indirect prompt injection (IPI) as a growing threat to AI systems and outlines Google’s continuous, layered defense approach—covering new attack discovery (human and automated red‑teaming, VRP, OSINT), synthetic data generation, deterministic and ML/LLM defenses, and Gemini model hardening—to detect, evaluate, and mitigate such attacks across Workspace apps.