Este curso contém o uso de inteligência artificial.

Google's Secure AI Framework (SAIF) is the practical guide to securing AI systems and agents throughout their entire lifecycle. This course transforms SAIF into something you can actually implement within a security or GRC program—without requiring ML engineering.

Working with a single realistic model organization, Meridian (a regulated digital banking institution launching a generative AI assistant for customers and then an agent flow), you'll go from first principles to a board-ready AI risk program.

What does the course cover?

• The six core elements of SAIF and why their order matters.

• The SAIF risk map — the four components: Data, Infrastructure, Model, and Application.

• All 15 SAIF risks — from data poisoning and model exfiltration to prompt injection, disclosure of sensitive data, and unauthorized actions.

• Controls mapped to risks, and how to create a control matrix for each risk.

• SAIF 2.0 — Protecting AI Agents: Human Oversight, Limited Powers, Observability, and the Agent Risk Map

• Operationalizing SAIF: risk self-assessment , remediation roadmaps, governance, RACI, and CoSAI.

• Mapping SAIF to NIST AI RMF, ISO/IEC 42001, OWASP LLM Top 10 and MITER ATLAS , in addition to reporting to the board and regulators.

Each section includes a quiz; six sections include solved tasks; and the course concludes with a complete final project — a comprehensive SAIF assessment, control plan, roadmap, and memorandum to the board for Meridian.

Who benefits: CISOs, security engineers, AI risk managers, and compliance and audit leaders who need to protect AI without waiting for a standard to dictate how.