When Performance Isn’t Enough: Ensuring the Safety and Security of AI Systems
Artificial intelligence
(AI) systems offer tremendous potential, but compared to traditional software,
they introduce novel safety and security risks. System theory provides a
powerful lens for understanding these risks and developing effective
mitigations. In this webcast, we’ll introduce System Theoretic Process
Analysis (STPA), a system-theory-based approach to safety analysis. We’ll
explain how STPA helps organizations build stronger assurances about the safety
and security of complex systems, including those that incorporate AI.
What Will
Attendees Learn?
- How complex systems fail due to design flaws and unsafe interactions—not just component failures
- How these types of accidents can occur in AI-enabled systems
- How to apply a system-theoretic perspective, including System Theoretic Process Analysis (STPA), to analyzing AI systems
- Practical insights into improving the design, testing, and operational use of AI systems to strengthen safety and security
Speaker Details
Dr. Matthew Walsh
Senior Data
Scientist within the CERT Division of Carnegie Mellon University’s Software
Engineering Institute (SEI). His research centers on leveraging artificial
intelligence (AI) to enhance cybersecurity and to usher in novel organizational
capabilities. Among his recent undertakings, Matthew has explored extensions of
system theory to assuring the safety and security of AI systems. Matthew earned
a Ph.D. in Cognitive Psychology from Carnegie Mellon University and a
bachelor’s degree in psychology from the Pennsylvania State University.
Dr. David Schulker
Senior Data Scientist and team lead in the Software
Engineering Institute’s (SEI’s) CERT Division, and his research focuses on
designing and applying machine learning and artificial intelligence (AI)
systems, testing and evaluating these systems, and analyzing their safety and
security. He has 13 years of experience in research supporting the Department
of War’s (DoW’s) operational and management challenges and is also a former Air
Force officer and a graduate of the U.S. Air Force Academy.
Event Topic
Artificial Intelligence, Machine Learning, SecurityRelevant Audiences
All Military, All Federal Government