In this document, Securboration proposes a Phase II effort to develop a technology called Zero-Trust Analytics for Machine Learning SecDevOps (ZTA-ML or Zeta ML). The overarching goal of the proposed effort will be to develop a zero-trust solution that enhances the security posture of operationally deployed ML software systems. ZTA-ML will be implemented as a dynamic analysis toolchain for ML software systems based upon Securborations existing Rampart product line that was originally developed for Java applications through non-SBIR research conducted by Securboration for DOT&Es Cybersecurity Assessment Program (CAP). ZTA-ML will track data movement and provenance, algorithm utilization, code-level behaviors, and platform-level interactions as they occur within running ML software systems. At runtime, ZTA-ML will identify and block abnormal sequences of these interactions (e.g., when some spurious input causes a hypothetical system to traverse a control flow pathway in a third-party ML library that was never previously executed). The current state of the underlying Rampart technology is well beyond the technical maturity required for a direct-to-Phase II topic. Technology developed under the proposed effort will have compelling commercial use cases and will therefore be a promising target for commercialization.