design, deploy, and govern a defense-in-depth architecture for AI workloads aligned with Google’s Secure AI Framework (SAIF).

Course introduction

The AI threat landscape

Data provenance, lineage, and copyright

Hardening the AI infrastructure

Guarding inputs and outputs with Model Armor

Agentic governance

Secure RAG pipelines

gsp2002-active-shielding-with-model-armor

Active Shielding with Model Armor

Course summary

Quiz: Security Best Practices in Google Cloud: Models — Question 1

Quiz: Security Best Practices in Google Cloud: Models — Question 2

Quiz: Security Best Practices in Google Cloud: Models — Question 3

Quiz: Security Best Practices in Google Cloud: Models — Question 4

Quiz: Security Best Practices in Google Cloud: Models — Question 5

Securing models

In this course, you'll learn how to apply Google's Secure AI Framework (SAIF) to autonomous agents, model integrity, and data pipelines.

<h2>Secure the Models Behind Your AI</h2>As organizations put AI models into production, those models become valuable assets and new targets for attack. This intermediate course, part of the Security Best Practices in Google Cloud series, focuses on protecting AI models and the systems around them.<br><br><h2>Apply Google’s Secure AI Framework</h2>You’ll learn how Google’s Secure AI Framework (SAIF) applies to models running on Google Cloud, and explore the main risks to AI models, including tampering, data poisoning, and misuse. You’ll see how to control who can access, deploy, and call your models, and how to protect the integrity of models and the data used to train them.<br><br><h2>Defend Models in Production</h2>The course then covers how to monitor and defend models once they’re serving real users, including guarding against harmful inputs and outputs. By the end, you’ll be able to apply security best practices across the AI model lifecycle on Google Cloud and fit model security into your wider cloud security strategy.

Manage Security in Google Cloud

Learn best practices for securing AI models on Google Cloud, from protecting model integrity to controlling access.

Active Shielding with Model Armor

Übungsanweisungen