Monica Mock-Sipos - The "Spy vs. Spy" Approach to AI Security in Banking | Red Hat

Host Grizz Griswold sits down with Monica Mock-Sipos, OpenShift Platform Leader for Financial Services (FSI) North America at Red Hat. Drawing on her background in enterprise IT architecture and cybersecurity, Monica explains why banks must shift their focus from perimeter "model security" to securing the full execution path of agentic AI workflows. From navigating non-deterministic logic and mid-workflow context shifts to deploying containment architectures on Kubernetes, this episode breaks down how financial institutions can safely operationalize AI without breaking compliance.🕒 Timestamps:* 0:00 - Welcome to the FINOS Podcast * 1:12 - What is an OpenShift Platform Leader? * 2:52 - Tech Sales vs. Architect-Led Engineering * 4:27 - From Building Architecture to Cyber Security * 7:04 - What Actually is Red Hat OpenShift? * 8:51 - Managing Non-Deterministic AI Risk in Finance * 10:11 - The Danger of Mid-Workflow Context Shifts * 11:31 - MCP Tool Exposures and the Anthropic Mythos Shock * 12:38 - Why You Must Secure the Full Execution Path * 14:42 - What to Expect at OSFF New York * 16:36 - "Spy vs. Spy": Why AI is the Only Defense Against AI 📊 The Problem: False Model Boundaries & Mid-Workflow Shifts* Inadequate Perimeter Defenses: Security strategies that only filter prompt inputs and outputs fail to protect the broader execution path where agents call external tools and internal banking APIs. * Mid-Workflow Context Shifts: Agentic workflows react dynamically. A transaction context can change halfway through a process—like a new fraud alert arriving during a delayed human approval step—derailing static security assumptions. * Vulnerability Proliferation: Custom DIY security stacks lack the bandwidth to keep pace with AI-accelerated vulnerability scanners, leaving unhardened environments open to machine-speed exploit chains. 🏗️ The Solution: Execution Path Hardening & Containment* Continuous Verification: Enforcing ongoing authentication, authorization checks, and least-privilege tool entitlements across every leg of a multi-step workflow until task completion. * Containment Architecture: Deploying containerized workload isolation on top of Kubernetes (OpenShift) so that runtime failures or compromised agent tasks cannot spread laterally into critical core systems. * AI vs. AI Defense: Embedding applied AI directly into the platform fabric to monitor, isolate, and remediate machine-speed attack vectors automatically (a "Spy vs. Spy" approach). ⚙️ Why This Matters for Financial Engineering* Safe Production Scaling: Empowers financial engineering teams to transition agentic AI prototypes into regulated production environments without breaching strict audit or compliance standards. * Audit-Ready Observability: Captures granular telemetry across all agent actions, tool calls, and state changes to deliver the verifiable, replayable audit logs required by financial regulators. 🌐 More about FINOS: https://www.finos.org/📧 Join our newsletter: https://www.finos.org/sign-up🎙️ Listen to our Open Source in Finance Podcast: https://www.youtube.com/@FINOS/podcastsLinkedIn: https://www.linkedin.com/company/finosfoundation