Home » AI Security Engineering: Defense, Strategy & Governance

AI Security Engineering: Defense, Strategy & Governance

Main Speaker

Tal Ben Shushan

Learning Tracks

Cyber Security

Course ID

42931

Date

23-06-2026

Time

Daily seminar
9:00-16:30

Location

Daniel Hotel, 60 Ramat Yam st. Herzliya

Overview

This comprehensive seminar provides a deep technical dive into the intersection of Artificial Intelligence and Cybersecurity. It explores the fundamental mechanics of modern AI architectures-such as LLMs, Transformers and Agentic workflows, while systematically analyzing the emerging threat landscape. Participants will learn how adversaries exploit AI through advanced techniques like prompt injection and data exfiltration, as well as how to implement robust governance and technical defense strategies using industry standards and secure infrastructure.

Who Should Attend

CISOs and Security Directors
Security Engineers and Practitioners
Technical Decision Makers and AI Architects
Developers and AppSec Specialists

Prerequisites

Course Contents

AI Fundamentals for Security Professionals

- Understanding the “AI Brain”: Neural networks, weights, parameters, and tokens.
- Modern Architectures: The Transformer revolution, attention mechanisms, and context windows.
- RAG (Retrieval-Augmented Generation): Semantic search, vector databases, and grounding.
- The Hardware Layer: CPU vs. GPU (CUDA cores), VRAM requirements, and performance monitoring.

The Technical Threat Landscape

- Prompt Injection: Direct attacks (Jailbreaking) and the high risk of Indirect Injection via external files (PDFs, Web).
- Data Leakage: Risks of training data exposure, mixing sensitive databases, and unauthorized API retrieval.
- Context Hijack: Sophisticated “hypnosis” techniques to override system instructions.
- Vulnerability Case Study: Analyzing real-world exploits like EchoLeak (CVE-2025-32711) targeting AI assistants.

Advanced AI-Powered Offensive Operations

- AI Social Engineering: Generating perfect phishing campaigns and identity fraud with Deepfake video/voice cloning.
- Agentic Attacks: Using autonomous agents (e.g., Cline) for multi-stage penetration testing and exploit execution via terminal orchestration.
- Tool Manipulation: Exploiting the MCP (Model Context Protocol) and Function Calling to gain OS access.

Technical Defense & AI Security Engineering

- Secure Prompting: Optimizing system prompts using XML delimiters and rigid security boundaries.
- AI Guardrails: Implementing real-time content filters, PII redaction, and hallucination controls.
- Observability & Auditability: Monitoring model invocations, latency, and cost using CloudWatch and CloudTrail.

Governance, Standards, and Deployment

- AI Governance: Implementing the ISO 42001 standard for Artificial Intelligence Management Systems (AIMS).
- Secure Infrastructure: Leveraging enterprise-grade environments (e.g., Amazon Bedrock) for isolated and encrypted model hosting.
- Deployment Strategies: Cloud-based APIs vs. local AI execution (Ollama/Llama) and their respective risk profiles.