September 29th | 1pm EST
SR 26-2 replaces the old checklist-based approach with materiality-based tiering, changing how banks validate models. It also names generative and agentic AI as out of scope for now, with an agency request for information reportedly on the way. That leaves model risk management leaders with the real work: deciding how much governance to build before the guidance catches up.
Join Nick Goble of Domino Data Lab for an open conversation with Jason DiPalma of KeyBank, Naresh Raheja, Former OCC Senior Risk Specialist and Examiner, and Lee Medoff, Former Head of Model and AI Governance at Fifth Third Bank on how model risk leaders are building governance and handling the generative and agentic AI gap.
The conversation will cover where agentic AI governance stands today and where supervisors and institutions are likely to take it next. Attendees will hear about:
- Classifying agentic systems: How to think about agentic systems that don't map cleanly onto traditional model risk categories, and why SR 26-2's silence on generative and agentic AI is a signal to keep building governance
- Monitoring autonomous agents: How oversight requirements shift as agentic AI moves from producing an output for a person to review to taking action on its own
- Interpreting the regulatory silence: A supervisor's perspective on how agentic AI will be examined, given SR 26-2 doesn't address it directly
- Governing what you didn't build: Why third-party AI risk is becoming as hard to govern as anything developed in-house
- Determining ownership: Whether agentic AI governance belongs in the first line, the second line, or a dedicated AI council, and how to make that decision
Comments