BoxVia

ANDEKS™ Normative System

ANDEKS™ provides a method and rules for independent assessment.

It helps reviewers explain what the evidence supports about an AI system, a digital product or an organizational process.

Read release v0.3 on Zenodo

ANDEKS™ in the TA-14 AI Governance Registry

ANDEKS™ is an independent assessment system that examines what the available evidence establishes about a specific AI model, version or configuration under declared conditions.

The TA-14 AI Governance Registry records ANDEKS™ v0.1 under TA-14-AIGR-000012, registered on .

The entry documents the declared scope of v0.1 and carries no certification, endorsement or technical validation.

View the TA-14 registry record

Read the TA-14 governance profile

How an assessment works

  1. Define what is being assessed and under which conditions.
  2. Check what the evidence supports and record what remains unestablished.
  3. Explain the conclusion so another reviewer can trace it back to the evidence.

The assessment informs a decision.

The decision remains with the person or organization responsible for it.

See the method in practice

Check a Claim is a small BoxVia demonstration of selected ANDEKS™ v0.3 principles.

It checks one claim against one public primary source.

A formal ANDEKS™ assessment follows the full applicable methodology.

Release v0.3

Published October 3, 2026.

The release includes four parts:

v0.1 was the first public release.

v0.2 was used internally in pilot work.

v0.3 is the second public release.

Zenodo hosts the English Controlled Derivative.

If the English and Russian texts differ in substance, the corresponding Russian source is authoritative.

Author and rights holder: Andrey Ekhmenin.

Comparing Independent Assessments

Results from different systems can be compared only with their questions, evidence and assessment boundaries in view.

Agreement on a result does not by itself establish that the systems or their authority are equivalent.

Related research explores how independent architectures preserve these distinctions.

Explore independent architectures and governance boundaries