PROBAITI Demonstrability Theory and Methodology Companion - Version 1.0
PROBAITI Demonstrability Theory and Methodology Companion - Version 1.0 provides the controlled explanatory companion to the PROBAITI Core Methodology Specification and develops the theoretical, methodological and operational reasoning underlying the PROBAITI approach to AI governance and assurance. PROBAITI is an evidence-first AI governance and assurance methodology concerned with what an organisation can demonstrably substantiate about a defined AI activity within an explicit boundary. It is designed to prevent governance conclusions from exceeding the authority, verified facts, controls, evidence, testing and continuing validity on which those conclusions depend. The Companion explains the principal PROBAITI mechanisms, including the Assurance Envelope, Demonstrability Chain, Evidence Ceiling Rule, Contradiction Rule, Demonstrability Debt, dependency-based change invalidation and targeted revalidation, dual traceability, and multi-axis non-compensation. It develops the methodology across authority architecture, claims, risks, controls and ownership; evidence and confidence; testing and challenge; contradiction handling; unresolved substantiation dependencies; lifecycle change; reconstruction; bounded conclusions; accountable decision-making; revalidation; competence and independence; methodology conformance; and controlled implementation. The Companion distinguishes PROBAITI propositions from external legal requirements. PROBAITI does not alter legislation, create a legal presumption of compliance, transfer regulatory authority or replace competent legal interpretation. Its methodological requirements become binding on an organisation only through an applicable governance instrument such as adoption, policy, contract or licence. The approved PROBAITI Core Methodology Specification - Version 1.0 is the normative methodological authority within the PROBAITI publication ecosystem. This Companion provides controlled explanation and interpretation within that boundary and does not override the Core Methodology. Version 1.0 was approved as a controlled PROBAITI methodology companion on 14 September 2026. This publication does not constitute legal advice, statutory conformity assessment, certification, accreditation, regulatory endorsement or independent empirical validation. Applicable external law and authoritative sources remain controlling within their respective scope.
Authors
- Cem Yılmaz
Institutions
- Centro Universitário Curitiba (BR)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-09-16
- DOI
- https://doi.org/10.5281/zenodo.22793489
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- article
- Field-Weighted Citation Impact
- 0.00