14-713 Special Topics: Secure and Responsible AI
12 units, Carnegie Mellon University
This course addresses the dual imperatives of security and responsibility in modern AI development. Students explore how to build AI systems that are resilient against attacks while respecting privacy, promoting fairness, and maintaining transparency. The curriculum integrates technical security measures with ethical frameworks, covering topics such as adversarial machine learning, differential privacy, fairness metrics, explainable AI, adversarial robustness, bias detection and mitigation, secure model training and deployment, and reliability assurance. Through hands-on projects, students develop AI solutions and then systematically evaluate their implementations across multiple dimensions: security vulnerabilities and defenses, reliability and performance under various conditions, ethical considerations including fairness and privacy, and overall trustworthiness for deployment. Students gain practical experience implementing security controls for machine learning pipelines, conducting fairness evaluations, assessing system reliability, and applying frameworks for building secure, ethical, reliable, and trustworthy AI systems. The course prepares students for careers developing AI systems in regulated industries and high-stakes applications where security, ethics, reliability, and trustworthiness are paramount. This course assumes students have some working knowledge of writing AI code and developing machine learning applications. A background in security (an Intro Course) and AI (an intro course) is preferred but not required.
Prerequisites: none listed
Offered: Fall 2026, Spring 2026
Courseweave needs JavaScript to show sections, workloads and planning tools.