SANS is the premium brand in security training, and SEC595 is its data science and machine learning course for security people. It runs for six days in person or live online, or as a self-paced OnDemand course with four months of access. Either way it costs $8,780, and the GIAC Machine Learning Engineer exam is an extra charge on top. The course is about applying ML to security, not about securing AI. You pull data from SQL and NoSQL sources, do statistics and exploratory analysis, build classical models and neural networks, and apply them to problems like finding anomalies in logs, hunting threats and classifying malware.
SANS says over 70% of the time is labs, and reviews from students agree the labs are the strength. The course expects intermediate Python but only pre-calculus maths, which makes it accessible to most analysts. It is a good course. The trouble is the price relative to what you can get elsewhere. The core ML content, how to train and evaluate a model, avoid leakage and pick features, is taught free and very well by Andrew Ng, fast.ai and others reviewed on this site.
What SEC595 adds is security datasets, a structured week and an instructor. Those have value, and for an analyst who would never finish a MOOC alone, a week off work in a SANS class may be what gets it done. One mismatch to be aware of: people searching for SANS AI security courses often want to learn how to attack or defend LLM systems. This is not that course. It teaches you to use ML as a tool. My take: excellent if someone else is paying and you need the structure.
Poor value if you are funding it yourself, because the free route plus public security datasets gets you most of the way.