Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Preprint . 2026
Data sources: ZENODO
addClaim

E-SPHERE Kernel 9.0 Interaction Governance Middleware for High-Risk Generative Artificial Intelligence (ISO-like Technical Specification)

Authors: Valente, Stefano;

E-SPHERE Kernel 9.0 Interaction Governance Middleware for High-Risk Generative Artificial Intelligence (ISO-like Technical Specification)

Abstract

E-SPHERE Kernel 9.0 is a deterministic interaction-governance middleware designed to support risk management and human oversight in high-risk generative artificial intelligence systems, in line with the requirements of the European Union Artificial Intelligence Act. The kernel operates independently of model architecture and content semantics, regulating instead the structural and temporal properties of AI–user interactions, including recursion intensity, interaction velocity, and semantic divergence. Governance actions are implemented through predefined state variables, deterministic escalation mechanisms, and auditable hard-stop conditions. The system explicitly excludes psychological assessment, behavioral profiling, or inference of human mental states. All indicators used are technical interaction metrics and are non-clinical by design. Validation was performed in a fully synthetic, pre-registered simulation environment comprising 1.5 million interaction cycles. Statistical evaluation using ANCOVA with orthogonalized covariates demonstrated high interaction stability (partial η² = 0.991, 95% CI [0.990, 0.992]), low residual variance, and no detectable covariance leakage. E-SPHERE Kernel 9.0 is intended as a governance-grade middleware to support compliance with Articles 9, 14, 15, and 52 of the EU AI Act and to contribute to future European standardization activities.

Keywords

High-Risk Artificial Intelligence Generative AI AI Interaction Governance Human Oversight Risk Management EU Artificial Intelligence Act Deterministic Safety Middleware Auditability AI Regulation European AI Standards

Powered by OpenAIRE graph
Found an issue? Give us feedback