Securing GenAI Systems: From Prompts to Autonomous Agents

A Hands-On Security & Architecture Course for Building Safe, Trustworthy, and Production-Ready GenAI Applications

What you’ll learn
Design secure GenAI architectures
Identify AI-specific vulnerabilities
Prevent prompt injection & data leakage
Secure agents & tool usage
Meet GenAI compliance requirements
Red-team and monitor AI systems

Requirements
Basic understanding of APIs, cloud services, and web security
Familiarity with LLMs (prompting, embeddings, RAG)

Description
Generative AI has changed how software is built — but it has also introduced entirely new security failures that traditional AppSec and cloud security models were never designed to handle.

This course is a deep, hands-on journey into the real security risks of modern GenAI systems, from prompt injection and RAG poisoning to tool abuse and autonomous agent failures. It is designed for software engineers, security engineers, architects, and AI practitioners who need to move beyond theory and understand how GenAI systems actually fail in production — and how to secure them properly.

Unlike high-level AI safety courses, this program is practical, adversarial, and systems-focused. You’ll break real GenAI workflows, observe emergent failures, and then implement concrete defenses using industry-aligned patterns.

By the end of this course, you won’t just understand GenAI security — you’ll know how to design, test, and govern AI systems safely at scale.

What You’ll Learn

Core Concepts

• Why GenAI security is fundamentally different from traditional AppSec

• How non-determinism breaks existing security assumptions

• Where trust boundaries actually exist in AI systems

• Why “prompt security” alone is insufficient

Hands-On Skills

• Exploit prompt injection and instruction hierarchy failures

• Poison RAG pipelines and observe real-world impact

• Abuse tool calling and function execution

• Trigger unintended behavior in multi-agent systems

• Implement real mitigations using policies, constraints, and governance

Defensive Architecture

• Secure RAG design patterns

• Tool and function authorization models

• Agent guardrails and bounded autonomy

• Policy enforcement outside the model

• Safe failure and human-in-the-loop design

What Makes This Course Different

• Hands-on labs, not slides

• Real failure modes, not hypothetical risks

• Agentic AI coverage (rare and critical)

• Security-first design mindset

• Aligned with OWASP LLM Top 10 & MAESTRO

• Built for production engineers, not researchers

Each week includes

• Conceptual video lessons

• Attack walkthroughs

• Jupyter-based labs

• Defensive redesigns

• Reflection and threat modeling exercises

Who this course is for
Software engineers building GenAI features
ML engineers & AI platform teams
Security engineers transitioning to AI security
Technical leaders & architects
Technical Product Managers