Episode 180: Gary McGraw on Machine Learning Security Risks

**Executive Summary:** This podcast transcript features Gary McGraw discussing the Berryville Institute of Machine Learning's architectural risk analysis of ML systems, highlighting 78 risks and a top-10 list (notably adversarial examples, data poisoning, transfer learning, data integrity/confidentiality, and reproducibility), emphasizing that data and model provenance are primary attack surfaces and urging engineers to "build security in" through risk-aware design and further mitigation research.