Program I
Responsible AI
Accountability · deployment · oversight · emerging technologies
Active research questions
Questions in development, with the current stage and intended publication format. Dates remain unannounced unless stated.
How do board risk committees develop AI risk appetite frameworks when system behavior cannot be fully specified in advance?
How should the three lines of defense be restructured for agentic systems, where ownership is distributed across product, engineering, compliance, and legal?
Where do MAS lifecycle controls, EU AI Act post-deployment obligations, FDA Total Product Lifecycle frameworks, and SR 11-7 evolution converge in operational expectations?
Broader lines of inquiry
- How do board risk committees set AI risk appetite when system behavior cannot be fully specified in advance?
- What does continuous lifecycle monitoring require operationally, and what is the build path for institutions starting from a compliance-only posture?
- How should the three lines of defense be restructured for agentic systems, where ownership is split across product, engineering, compliance, and legal?
- Where do MAS, EU AI Act, FDA, and SR 11-7 obligations actually converge in operational terms?
Image credits
- Responsible AI
- Human/robot teaming. Image: S. Bagchi/NIST.