AI Proof Layer: An Outcome-Assurance Architecture for Reliable, Safe, and Auditable AI Models

Read full text Summarise with VIRAT
Authors: Patrick Casimir
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Large Language Models (LLMs) deliver exceptional generative capability, but they share a structural limitation: they cannot prove that a given output is correct, grounded in authoritative evidence, or compliant with policy. As a result, hallucinations, unverifiable claims, and policy violations persist, especially in high-risk settings such as finance, healthcare, legal reasoning, and enterprise operations. This paper introduces AI Proof Layer, an external outcome-assurance layer that operates independently of the model. AI Proof Layer evaluates model outputs against explicit, measurable guarantees (“claims”), enforces ALLOW/BLOCK decisions, and generates immutable Evidence Packs suitable for audits, incident reviews, and regulatory reporting, without modifying or constraining the underlying model architecture. By separating generation (the model) from permission (AI Proof Layer), the system converts generative AI models from a probabilistic text generator into a certifiable decision system. We present the conceptual framework, reference architecture, decision contract, example workflows, and compliance mappings that enable organizations to reduce hallucinations and establish traceable accountability across the AI lifecycle.

Fields of Study & Subjects

Explainable Artificial Intelligence (XAI) Ethics and Social Impacts of AI Safety Systems Engineering in Autonomy

Publication details

Journal / source
Journal of Artificial Intelligence & Cloud Computing
Volume
5
Issue
2
Pages
1–1
Publication date
4 April 2026
Language
EN