SentraMind

Primer

AI-security primers

Short reference notes on detection order, evidence, checkpoints, model absence, and air-gap language.

Primer

Deterministic first, then anything else

Why a governed AI-security decision starts with a byte cap, a parse, and a deterministic detector before any model-shaped step.

Primer

Evidence is ciphertext at rest

What the two-level evidence envelope is for, what the database is allowed to hold, and what this POC does not prove about hardware custody.

Primer

What an audit-chain checkpoint proves

A checkpoint signs a record naming the chain head. It does not, by itself, prove that history was not rewritten, that a SIEM accepted it, or that a human reviewed the incident.

Primer

Model absence is a state, not a fallback

When a model-shaped task has no configured provider, the platform reports that state. It does not silently swap in another model.

Primer

What air-gap-capable means on this POC

The appliance is designed for air-gapped operation. Runtime zero-egress is an evaluation objective, not a measured proof in this build.