AI Governance Platform journal
Ideas for systems that make decisions.
Notes on the architecture, protocols and operating principles behind trustworthy AI systems.
Featured writing
From the control plane
Stateless MCP did not require a control-plane rewrite
Keycloak already gave us request-scoped identity and our governance state was explicit and durable. We still upgraded the MCP transport for compatibility, scale, and simpler gateway operations.
Build with clear boundaries
AI Governance Platform separates governance state from runtime execution, making policy and evidence easier to reason about.
Protocols are interfaces
The right transport depends on who—or what—needs to consume the capability.
AI agent skills are not system knowledge
Reusable agent skills teach workflows. Repository instructions capture the architecture, ownership, and non-negotiable rules that make a specific system safe to change.
Lineage is not analytics
A graph should make governed relationships inspectable. It should not pretend to be a cost dashboard, a quality ranking, or an automated recommendation engine.
Why dataset metadata belongs in the control plane—and dataset bytes do not
Immutable datasets need both a governed registry record and an artifact store. Combining them produces a system that is harder to scale, audit, and replace.
Observed is not managed: the ownership model AI registries miss
A prompt or model in a governance system is often execution evidence, not an artifact authored by the platform. Treating those states differently changes trust, privacy, and integration design.
A governed asset plane for AI systems: catalog first, MLOps later
AI Governance Platform now records versioned prompts, models, datasets, and evaluation providers with explicit ownership, immutable artifacts, and ontology-backed lineage.
Replay is not retry: one word that changes the integrity of the evidence
A retry asks whether work eventually finished. A replay asks whether a governed system can reproduce and defend a historical decision.
The incident report is not evidence
When an AI decision matters, logs and dashboards are not enough. The system must preserve the evidence required to reconstruct the decision itself.
The hard scaling problem in AI is deciding what to remember
Millions of historical executions do not call for a bigger dropdown. They call for a clear boundary between search projections and authoritative evidence.
MCPO vs Streamable HTTP: choosing the right interface for AI Governance Platform
MCP transports and REST adapters solve different integration problems. Here is how to choose between them—and why AI Governance Platform can support both.