Humans and Trustworthy AI Systems

Registration:

---

---

About

Many of the central challenges of trustworthy AI arise at the intersection of technical systems and human behavior. The seminar is jointly taught and is designed around an approximately equal balance between human-centered usable security and privacy topics and the technical security of AI, machine-learning, and large language model systems. Further information about presentation formats, topic assingment, and assessment requirements will be provided at the beginning of the seminar.

---

Prerequisites

This seminar is aimed at master's students in Cybersecurity or related fields. Participants should have a basic understanding of computer security, privacy, artificial intelligence, and machine learning and an interest in examining these technologies from both technical and human-centered perspectives.

---

Learning Objectives

Students will develop a critical understanding of how AI systems affect security, privacy, fairness, and society. They will learn to analyze trustworthy AI systems from complementary human-centered and technical perspectives, assess relevant research, identify limitations and potential harms, and communicate complex findings clearly. Particular attention will be paid to the gap between an AI system's intended behavior and the ways in which it may be understood, used, misused, or experienced in practice.

---

Seminar Content

The seminar explores trustworthy AI through an equal combination of two perspectives.

The human-centered perspective focuses on usable security and privacy, including

• users' mental models of AI,
• privacy concerns,
• vulnerable or marginalized groups,
• technology-facilitated abuse, scams, and misinformation.

The technical perspective examines the security of AI, machine-learning, and large language model systems, including

• secure code generation,
• the offensive use of AI,
• risks introduced by autonomous agents, and
• methods for detecting generated media.

Students will engage with recent scientific publications, present one selected topic, and lead a critical discussion of its human and technical dimensions. Presentations should explain the underlying technology, examine relevant security and privacy concerns, consider how people understand and interact with the system, and discuss how proposed safeguards may fail in real-world settings.

---

Special Notes

This seminar does not include a written seminar paper. In light of the growing use of generative AI in writing, the course instead places particular emphasis on students' own understanding, presentation, critical reasoning, and active participation in discussion. Assessment will therefore be based on presentations and contributions to seminar discussions.

---

Talk Schedule (Tentative)

---

Time and Location