Securing Agentic AI: Threat Modeling Autonomous Systems

May 23, 2026 | 9:00am - 3:00pm

As AI systems become more agentic, autonomous, and tool-capable, they introduce a new class of security risks that traditional models were not designed to address. Unlike passive AI systems, agentic systems can retrieve data, call APIs, make decisions, and trigger downstream actions, giving attackers more opportunities to manipulate behavior, abuse privileges, and compromise trust.

This hands-on, beginner-friendly workshop explores how the attack surface expands and how these threats appear in real-world environments.

In this session, you’ll:

Understand what agentic AI is and why autonomy introduces new security risks
Learn how attackers target AI systems in practice, not just in theory
Break down an agentic AI solution into clearly attackable components

Explore real‑world threats, including:

Prompt injection
Tool & API abuse
Memory and context poisoning
Over‑privileged AI agents
Apply basic security and governance controls aligned with OWASP and NIST
Participate in a guided group exercise to build a beginner‑level AI threat model
Engage directly with an industry practitioner on how AI is secured in real environments

You’ll leave with:

A strong foundation in agentic AI security risk concepts
Hands‑on experience building an adversarial AI threat model
Practical knowledge applicable to coursework, internships, and entry‑level roles
Career insights and expectations for roles in AI security, engineering, and GRC

Who Should Attend

University students (Cybersecurity, Computer Science, Data, IT, Risk, GRC)
Early‑career professionals entering cybersecurity, AI, or risk roles
Anyone curious about real‑world AI security beyond theory

Register

Special offer! The first 30 UCSC students, staff, and instructors to register can attend for free with the code: UCSC2026.

Claim your seat.

No prior AI security experience required. Please bring your own laptop and charger. Light breakfast and lunch will be provided.