Overview
AI governance, in plain terms.
AI governance is the set of practices that make sure an AI system is safe, fair, explainable, and accountable across its whole life — from the data it learns on to the decisions it makes in production. It is not a compliance checkbox; it is how an organization stays in command of systems that now make or shape real decisions about real people.
Why financial services is different
In financial services, AI doesn’t sit in a sandbox — it decides who gets credit, flags who might be committing fraud, screens who can open an account, and drafts what customers are told. Each of those touches a regulated outcome, and each carries a cost when it goes wrong: unfair denials, missed fraud, discriminatory patterns, or advice that misleads. The regulatory expectation is that you can explain a decision, evidence how the model was validated, and name who is accountable for it.
Where the risk lives
Fairness & bias
Does the model treat protected groups equitably across its decisions?
Explainability
Can you explain a decision to a customer, an auditor, and a regulator — at the level each requires?
Data lineage & privacy
Do you know where the training and input data came from, and are you using only what you need?
Reliability & drift
Does the model still behave in production the way it did when it was validated?
The frameworks I work against
Good governance connects technical capability to regulatory expectation. The frameworks I most often map to: the NIST AI Risk Management Framework (Govern, Map, Measure, Manage), the EU AI Act (risk tiers, data governance, transparency, and human oversight), SR 11-7 / model risk management (the long-standing US banking standard), and sector rules on fair lending and consumer protection. The goal isn’t to memorize them — it’s to translate them into controls a product team can actually build against.