Skip to content
BENIBIX Learn

Model Risk Management (MRM)

Business🏛️ Governance & EthicsCompliance and Risk

Also known as MRM, Model Risk, SR 11-7, SR 26-2

Checking that the models a firm uses to decide are right. And that they are used right.

Think of it like

It is like a second cook tasting the soup. One cook makes it, and the other checks it first.

Example

In April 2026 the US Federal Reserve, the FDIC and the OCC replaced their 2011 guidance, SR 11-7, with a new one, SR 26-2. It matters most for banks with over 30 billion dollars in assets.

Why it matters

A wrong model can make thousands of bad choices before anyone notices. A second check catches the error early.

Definition

Model risk management is how a firm controls the harm from models that are wrong or used wrongly. A model here is any method that turns data into a number for a choice, like a credit score. Each model is listed and tested by someone who did not build it. It is then watched over time. US banks have had formal rules for this since 2011, and now apply them to AI too. The key idea is effective challenge, a real and informed second look.
The team that builds a model is the first line. A second line tests it, and audit checks both.

Common mistake

Letting the team that built a model also approve it. The check only works if it comes from someone else.

Key takeaways

  1. 1
    Full list. Know every model in use.
  2. 2
    Independent test. Someone else tests each model.
  3. 3
    Keep watching. Models drift as the world changes.