VVZ API is not affiliated with ETH Zurich. Data might be outdated or incorrect. Please view the official ETHZ Vorlesungsverzeichnis for binding information.
AI Security Seminar: From Code to Agent Security
Last Updated: 2026-06-03 00:07:36
Abstract
The seminar covers various topics at the intersection of AI and security, including AI for code security and the security of AI agents.
Objective
The objective of the seminar is to: - Introduce students to the fields of AI security and AI for security. - Learn practical techniques on how to attack and secure code and agentic systems using Large Language Models (LLMs). - Highlight the latest research and work opportunities in industry and academia available on these topics.
Content
The rapid integration of AI into the software development lifecycle has created a paradigm shift in security. This seminar explores two critical frontiers: AI for Code Security and the Security of AI Agents. As AI models become proficient at generating and reasoning about code, they offer unprecedented opportunities to automate "hard" security tasks. We will examine how well LLMs can generate secure code and how they facilitate automated vulnerability discovery, and the synthesis of secure patches. Unlike traditional static analysis, these AI-driven approaches can navigate the vast complexity of modern software repositories to identify subtle logic flaws that previously required manual auditing. Simultaneously, the rise of LLM based AI Agents introduces new security vulnerabilities. Because these agents operate on natural language interfaces, with no clear separation of instructions and data, they are susceptible to various attacks. We aim to understand how to attack and secure those agents. In this seminar, we will cover the latest developments in these fields, including adversarial robustness, formal verification, and secure architectures for code and agentic workflows. The seminar is carried out as a set of presentations chosen from a list of available papers. The grade is determined as a function of the presentation, handling questions and answers, and active participation in discussions.
General Information
- Language
- English
- Levels
- MSC
- Frequency
- Yearly recurring
Examination
- Type
- graded semester performance
Registration & Places
- Max Places
- 22
- Signup End
- 28.09.2026
Course Components
| Type | Title | Time & Place | Hours |
|---|---|---|---|
| seminar | AI Security Seminar: From Code to Agent Security | No time listed | 2 h weekly |