Intelligent Governance for the AI-Driven Enterprise
Insights/Technical

Model risk management in the age of generative AI

Extending SR 11-7 thinking to foundation models, prompts and retrieval pipelines.

DK
Dr. Kabir Shah
Head of Model Risk · 2026-04-24 · 1 min read
inX🔗
Technical
Illustrative — drop in a chart, diagram or photo for the live article.

SR 11-7's principles — sound development, effective validation, and robust governance — still apply to generative systems. What changes is the surface area: prompts, retrieval sources, and non-deterministic outputs.

Validating a moving target

Traditional validation assumes a stable input-output mapping. With foundation models, validation must cover prompt templates, retrieval quality, and guardrail behaviour under adversarial input.

Model RiskGenerative AISR 11-7
DK
Written by
Dr. Kabir Shah

Head of Model Risk at MS RiskTec. Dr. advises boards and CROs across BFSI on model risk, explainability and regulatory readiness for AI-driven lending.

Work with our advisory team →

Turn FREE-AI from a burden into an operating model

Start with an AI risk assessment and see where your credit governance stands today.