Hardening MCP: Advanced Security and Scale for Enterprise AI Hubs | THR978
EY Kalman presents the production architecture and security controls needed when MCP servers become part of enterprise AI data pipelines. The session focuses on practical deployment patterns and the boundaries that must hold as usage grows.
Architecture and secure request handling
Drawing on real-world FastTrack deployments, the session examines the architectural implications of deploying the stateless MCP standard. Key topics include:
- Stateless scaling: Load balancing and gateway patterns for enterprise MCP deployments.
- Multi-tenant routing: Routing requests while maintaining tenant security and corporate data boundaries.
- Authentication: The evolution of MCP authentication and secure request handling.
- Least privilege and trust validation: Applying these principles within a production architecture.
Security and operational governance
Kalman addresses prompt-injection risks alongside zero-trust AI security. The discussion also covers the operational controls needed to manage enterprise exposure:
- Governance, monitoring, logging, and threat reporting.
- Enterprise risk management, expiring access, and network isolation.
- Throttling and escalation to human oversight.
- Governance practices for securely scaling AI data pipelines.
The session concludes with a Microsoft Azure reference architecture and lessons from real-world MCP security deployments.
Session chapters
- 00:00 — Introduction to MCP and enterprise-scale challenges
- 01:11 — Scaling AI systems and preparing for security threats
- 02:00 — Production architecture, least privilege, and trust validation
- 03:03 — MCP authentication evolution and secure request handling
- 04:49 — Load balancing, gateways, and multi-tenant security patterns
- 08:58 — Governance, monitoring, logging, and threat reporting
- 12:01 — Enterprise risk management, expiring access, and network isolation
- 14:43 — Prompt-injection risks and zero-trust AI security
- 17:08 — Throttling, human escalation, and governance best practices
- 20:34 — Microsoft Azure reference architecture and real-world MCP security lessons