rAIson by Argument Theory

Compliance

Compliant by design — not bolted on.

The laws now governing AI risk demand transparency, human oversight, and traceable rationale. A named-rule trace and identifier-free inputs are exactly what rAIson already produces.

What the architecture already does

Oversight, explanation, and reproducibility are structural — not features.

Human oversight

An expert reviews the encoding unit by unit before deployment — the model proposes, the expert approves, then the engine commits. Errors are caught in the source, not the consequences.

Intelligible explanation

Every decision is the rule trace itself: the rules that fired and the counter-rules they defeated. The explanation is not generated after the fact — it is the reasoning.

Identifier-free inputs

The engine reasons over structured signs and codes, never personal identifiers — keeping decisions auditable while minimising the data they touch.

The regimes

Mapped to the laws governing high-risk AI.

EU AI Act · Articles 13 & 14

Transparency and human oversight

High-risk AI must give an intelligible explanation for every output, and a human must be able to review and override every decision. A black box cannot satisfy this; a named-rule trace can.

FDA SaMD · 2025 AI guidance

Lifecycle management of medical software

Predictable behaviour, documented updates, and a traceable rationale for every clinical recommendation. Editable rules and signed updates beat opaque retraining cycles.

GDPR · Articles 4 & 22

Automated decisions and the right to explanation

The architecture transmits only structured signs and codes — never identifiers — and produces explanations that are the rule trace itself.

EU MDR Class IIa · HIPAA

Medical devices and protected health information

Quality-management systems, post-market surveillance, and Safe-Harbor-grade de-identification — delivered by auditable updates and identifier-free inputs.

Compliance is not an add-on layer. It is what the architecture was already doing to produce reproducible, explainable, defensible decisions.

Decisions you can take to an auditor.

See how a named-rule trace satisfies the regimes that apply to your domain.

Request access