ANDEKS™

ANDEKS™ is an independent governance assessment system for evaluating a specifically identified AI model, version, or other distinguishable configuration against explicitly declared properties, claims, conditions of use, boundaries, and available evidence. It is intended for organizations and other decision-makers that need a bounded, attributable assessment before relying on broader governance, assurance, regulatory, or operational conclusions. ANDEKS™ preserves the identity of the assessed object, the declared subject of assessment, the evidentiary basis, and the limits of the resulting finding. It does not certify, license, approve, authorize, or govern execution, and it does not transfer responsibility for downstream decisions to the assessor or to ANDEKS™.

ANDEKS™ is stewarded by **Andrey Ekhmenin**.

The registered architecture identifies ANDEKS™ as an independent governance assessment system intended for organizations and decision-makers that need a bounded and attributable assessment of a specifically identified AI model, version, or distinguishable configuration before relying on broader governance, assurance, regulatory, or operational conclusions.

The TA-14 Registry presently records Andrey Ekhmenin as the steward of ANDEKS™ version 0.1. No organization is separately declared in the Registry baseline.

ANDEKS™ is building a governance assessment architecture centered on **bounded object identity, declared assessment scope, evidence, and attributable findings**.

Its model requires the object being assessed to remain specifically identified rather than allowing a finding about one version, configuration, deployment, or evidentiary state to silently expand into a broader claim about an entire model family, provider, platform, or operational environment.

The architecture also preserves versioned normative baselines so changes to ANDEKS™ itself can be distinguished from earlier released states.

This makes ANDEKS™ primarily an **assessment-governance architecture**, not an execution-governance system.

In its registered declaration, ANDEKS™ states that it provides a normative framework for independent assessment of a specifically identified AI model, version, or distinguishable configuration against explicitly declared properties, claims, conditions of use, boundaries, and available evidence.

The registered architecture requires both the assessed object and the subject of assessment to remain explicitly identified and bounded. Findings are constrained to the fixed object, declared subject, applied conditions, and evidentiary basis used during the assessment.

ANDEKS™ also declares versioned normative baselines and attributable release lineage.

Its registered non-claims are equally important. ANDEKS™ does not claim to certify, license, accredit, approve, or authorize an AI system or organization. It does not establish legal or regulatory compliance, provide a universal model-safety determination, govern runtime execution, make deployment decisions, or transfer downstream responsibility to the assessor or to ANDEKS™.

TA-14 is paying attention to ANDEKS™ because it preserves several distinctions that matter in serious AI governance: **what object is actually being assessed, what evidence supports the assessment, what conditions apply, and how far a resulting finding may legitimately travel**.

Those boundaries matter because an assessment can become misleading when a finding about one model version, configuration, evidence set, or operating condition is generalized beyond the object actually examined.

ANDEKS™ addresses that problem at the assessment layer by keeping object identity, declared scope, evidence, and finding boundaries attributable.

TA-14 remains a separate architecture concerned with admissible execution across the broader consequence-bearing chain. ANDEKS™ does not claim that execution-governance role, and this profile does not assign it one.

Registration establishes the attributable baseline for this governance architecture in the TA-14 AI Governance Registry. It does not certify the architecture, endorse its technical claims, determine legal validity, establish regulatory approval, adjudicate ownership, or prove technical performance.

Future governed work may include additional ANDEKS™ versions, demonstrations, independent reviews, evidence challenges, technical comments, participant responses, or related execution artifacts where separately submitted and governed.

Any future TA-14 review should remain tied to a specifically identified ANDEKS™ version, declared assessment claim, evidence package, and review scope.

Later ANDEKS™ versions should remain separately attributable so the Registry can preserve lineage rather than silently replacing earlier states.

ANDEKS™ matters because it treats the **boundary of an assessment itself as a governance problem**.

An AI governance finding is only as trustworthy as the identity of the object examined, the evidence admitted, the conditions applied, and the limits placed on the resulting conclusion. When those elements drift, a bounded assessment can become an unbounded claim.

ANDEKS™ addresses that risk by preserving a distinguishable assessed object, explicit assessment subject, evidentiary basis, conditions, limitations, and version lineage.

From the TA-14 institutional perspective, that makes ANDEKS™ a useful example of an architecture occupying a distinct governance layer: **independent assessment before broader reliance**, while expressly declining to treat assessment as certification, authorization, regulatory approval, or execution governance.