AISYSTEMS Architecture

The architecture of AISYSTEMS is designed as a multi-layered model with strict separation of responsibilities, authorities, and operational domains.

Its objective is to ensure audit neutrality, technical integrity of execution, and long-term stability of reference layers.


AISYSTEMS s. r. o.
│
├── Governance Layer
│
├── Secure Infrastructure Layer
│
├── Audit Authority
│   └── Audit Framework (observe-only)
│
├── Execution Runtime
│   └── AISYSTEMS SENTINEL
│
├── Reference Layer
│   └── AI Core (public reference ontology)
│
└── Application Layer
    └── Independent AI services (separate application cores)

Layer Description

Governance Layer

Defines ownership structure, versioning rules, responsibilities, and delegation of authority. It does not interfere with audit evidence and has no exemption from methodological neutrality.

Secure Infrastructure Layer

Operational layer ensuring runtime isolation, execution integrity, and separation of control planes. It does not interpret audit results.

Audit Authority — Audit Framework (observe-only)

A deterministic, non-intervening methodological layer designed for technical observation of AI system behavior.

The Audit Framework:

  • does not modify configuration of target systems
  • does not optimize their behavior
  • generates exclusively structured technical evidence
  • does not issue evaluations, certifications, or verdicts

It defines methodology and audit rules. It does not perform execution.

Execution Runtime — AISYSTEMS SENTINEL

AISYSTEMS SENTINEL is the deterministic runtime implementation of the Audit Framework.

Operating in observe-only mode, it executes an AuditPlan, interacts with target systems as a black box, and produces cryptographically verifiable technical artifacts.

It has no authority to modify methodology or audit rules. It is an execution layer, not an authority.

Reference Layer — AI Core

A public reference ontology stabilizing terminology used in technical observation of AI systems.

AI Core:

  • does not prescribe implementation
  • is not an audit tool
  • does not provide certification

Application Layer

Independent application solutions and AI services architecturally separated from the audit layer.

These systems have no privileged access to audit methodology.


Architectural Principle

Methodology, measurement, reference terminology, and application services remain strictly separated.

Audit produces evidence. Decision-making belongs to external entities.