01
Agentic Responsible AI Compendium
Financial institutions have spent four decades building a mature discipline, model risk management in the SR 11-7 tradition, for governing systems that estimate.
Research library
Evidence,
made operational.
Primary sources, state-of-the-science synthesis, control frameworks, and serious analysis selected for implementation consequence.
Original reports
From accumulated evidence to durable guidance.
These reports are now readable as stable web editions; available source PDFs remain downloadable from each record.
02
Agentic Measurement & Evaluation: Must-Have Meta-Analysis, 2023–2026
Agentic evaluation has matured from "can the model answer?" into a layered assurance stack for systems that plan, use tools, persist across steps, and can affect external state. The must-have eval program is not one benchmark. It is a portfolio:
03
Agentic RAI Controls and Autonomous Workload Safety
Most of the depth in agentic safety research answers a lab's question, is this model safe enough to release?, through capability evaluation and adversarial control protocols.
04
Enterprise Assurance Stack: What’s New
Five primary sources arrive together and, read as a set, they are not five loose standards. They are a layered AI-assurance stack that maps cleanly onto the classic three-lines-of-defense governance model.
Current evidence shelf
Selected primary research
Agent Safety Is Action Alignment
Reframes agent safety around authority and execution pathways.
Agentic systemsGovernance
On-Policy Self-Evolution via Failure Trajectories
Evaluates repair across complete failure trajectories.
Agentic systemsEvaluation
The Hot Mess of AI: Misalignment Scaling Verification
Tests whether longer reasoning produces coherent or chaotic misalignment.
Agentic systemsGovernance
Model AI Governance Framework for Agentic AI v1.5
Translates agent governance into runtime bounds and intervention procedures.
Agentic systemsGovernance
Guidelines on transparency obligations
Defines current transparency obligations for providers and deployers.
Agentic systemsGovernance