Model risk management focuses specifically on the risks that come from AI and machine learning models themselves - the possibility that a model is wrong, biassed, or behaving in ways you don't expect. This discipline has deep roots in financial services, where statistical models have been subject to rigorous governance for decades, but it's increasingly relevant across all industries as AI models make more consequential decisions. Effective model risk management includes validation before deployment (testing the model thoroughly with representative data), ongoing monitoring (tracking performance metrics and watching for drift), documentation (recording how the model was built, what data it was trained on, and what its known limitations are), and periodic review (reassessing whether the model still performs acceptably). The complexity depends on the stakes: a model that suggests which articles to read requires lighter governance than one that influences medical diagnoses or credit decisions. But even low-stakes models deserve basic hygiene - knowing what's deployed, how it's performing, and who's responsible for it. Without this foundation, you can't manage what you don't know you have.