| |
 | | Towards Operationalizing Technical AI Governance |  | Jose Luis LUNA CAMPOVERDE PhD Candidate School of Computing and Information Systems Singapore Management University FULL PROFILE |
Research Area - Artificial Intelligence & Data Science
- Machine Learning & Intelligence
Dissertation Committee | Advisor: | | | Co-Advisor: | | | Members: | |
| | Date 25 November 2026 (Wednesday) Time 9:00am – 10:00am Venue Meeting room 5.1, Level 5 School of Computing and Information Systems 1, Singapore Management University, 80 Stamford Road, Singapore 178902 Please register by 23 November 2026. We look forward to seeing you at this research seminar.
|
| ABOUT THE TALK Generative and agentic AI systems are being deployed faster than the rules meant to govern them. Regulations, standards and government frameworks increasingly say what these systems should achieve: robustness, fairness, security across the AI supply chain, and meaningful human oversight. They rarely say how to turn those expectations into technical requirements, controls and evidence that can be checked in practice. This dissertation proposal addresses that gap. It argues that governance expectations must be translated into technical properties that fit the deployment context and can be traced back to the requirements they come from. These properties should be grounded in where risks arise in real systems and in how organizations actually encounter and control them.
The proposal develops this argument in stages. It first establishes what governance instruments actually expect of generative AI, by comparing government frameworks across regions. It then traces where security and safety risks originate in real systems, mapping threats to specific components of the large language model supply chain. This work informed Singapore’s Technical Reference TR 147:2026. Next, it examines how enterprises deploying agentic AI encounter and control these risks in practice. It finds that current safeguards focus on constraining agents rather than on assuring, containing and recovering from their actions, and that existing guidance is too high-level to apply directly. It also shows that deployment context should shape how systems are assessed, so that the evidence required scales with the risk of the use case. The proposed final study brings these threads together. It turns guidance clauses into testable properties and executable test cases for agentic systems, starting with the insurance sector. Overall, the work moves technical AI governance from stating what is expected towards producing evidence of whether it is met. | ABOUT THE SPEAKER Jose Luis LUNA CAMPOVERDE is a PhD candidate at the School of Computing and Information Systems, Singapore Management University, under the supervision of Assistant Professor Xiaofei XIE. Jose’s research is in technical AI governance, focusing on how AI regulatory and standards requirements can be translated into technical practice for generative and agentic AI systems. In addition to his academic research, he has contributed to national AI standardization efforts, including his role as co-convener for Singapore’s TR 147:2026 on cybersecurity practices for LLM applications. |
|