AI Security & Governance
Frameworks and practical guidance for securing and governing AI systems.
- Basic Idea of AI Governance
The governance lifecycle, agent threats, layered controls, and what a governed AI system needs on record before it ships.
- AI Identity & Criticality Tiers
People, AI agents, and APIs on one shared risk ladder — how authentication and review cadence should scale with what’s at stake.
- Azure AI Governance
A 360° view of how Azure Policy, Microsoft Foundry, Cognitive Services, and Microsoft Purview fit together to govern AI in practice.
- AWS AI Governance
The same governance shape rebuilt from AWS primitives — SCPs, Bedrock, SageMaker, Macie, and Security Hub — mapped honestly against where it diverges from Azure.
- Securing Agentic AI Systems
Where risk concentrates once agents chain into each other — confused deputy, cascading failure, and a secure design-to-operate lifecycle for multi-agent systems.
- AI Agent Memory — Types & Security
The kinds of memory an agent keeps — episodic, semantic, procedural, reflective, and by structure (vector, key-value, graph, summary, external) — and why every one is also an attack surface: poisoning, persisted injection, cross-tenant leakage, retention.
- Govern, Map, Measure, Manage
The NIST AI Risk Management Framework end to end — the four functions, the seven characteristics of trustworthy AI, how it compares to ISO/IEC 42001 and the EU AI Act, and where programs actually fail to operationalize it.
- Defense in Depth for Bedrock AI Chat
A reference architecture for AI chat features that can’t run away in cost — eleven independent layers, why each one is separate, and the actual code behind them.