There is a newer version of the record available.

Published April 17, 2026 | Version 1.0
Preprint Open

Non-Invertible External Supervisory Control: A Theoretical and Architectural Framework for External Supervision and Explicit Operational Risk Management in Large-Scale Artificial Intelligence Systems

Authors/Creators

  • 1. Independent Researcher

Description

This preprint introduces the Non-Invertible External Supervisory Control (NIESC/CSENI) framework, a theoretical and architectural proposal aimed at examining the structural limitations of AI safety approaches based exclusively on internal alignment. The paper argues that, under conditions of model opacity, limited auditability, and strategic capability of the supervised system, safety cannot rely solely on restrictions embedded within the same computational substrate as the agent.

CSENI is formulated as an external supervisory layer centered on three elements: controller decoupling, gradual operational friction, and heterogeneous observables. The paper includes a preliminary threat model, a minimal formalization of the supervisor, an initial experimental agenda, and a technical appendix with a reproducible minimal instantiation for tool-using agents.

This work does not claim to offer a definitive solution to the control problem for advanced AI systems. Its purpose is foundational: to propose a conceptual vocabulary, a minimal formal basis, and a forward-looking program for future technical, regulatory, and institutional developments.

Abstract (Spanish)

Este preprint presenta el marco de Control Supervisado Externo No Invertible (CSENI), una propuesta teórica y arquitectónica orientada a examinar las limitaciones estructurales de los enfoques de seguridad en inteligencia artificial basados exclusivamente en alineación interna. El trabajo argumenta que, bajo condiciones de opacidad del modelo, auditabilidad limitada y capacidad estratégica del sistema supervisado, la seguridad no puede descansar únicamente en restricciones embebidas dentro del mismo sustrato computacional del agente.

CSENI se formula como una capa de supervisión externa basada en tres elementos centrales: desacoplamiento del controlador, fricción operativa gradual y uso de observables heterogéneos. El texto incluye un modelo de amenaza preliminar, una formalización mínima del supervisor, una agenda experimental inicial y un apéndice técnico con una instanciación mínima reproducible para agentes con herramientas.

Este trabajo no pretende ofrecer una solución definitiva al problema del control de sistemas avanzados de IA. Su propósito es fundacional: proponer un lenguaje conceptual, una base formal mínima y un programa de trabajo posterior para desarrollos técnicos, regulatorios e institucionales más robustos.

Files

CSENI_Marco_Control_Supervisado_Externo_No_Invertible_v1.0_ES_equations_fixed.pdf

Files (303.3 kB)

Additional details

Additional titles

Alternative title (Spanish)
Control Supervisado Externo No Invertible: Un Marco Teórico y Arquitectónico para la Supervisión Externa y la Gestión Explícita del Riesgo Operativo en Sistemas de Inteligencia Artificial a Gran Escala

Related works

Is supplement to
Preprint: 10.5281/zenodo.17702378 (DOI)
Preprint: 10.5281/zenodo.19471413 (DOI)

Dates

Issued
2026-04-17
First public preprint release

References

  • Rivera-García, J. M. (2025). Teoría Unificada de la Inteligencia (v4.2): Marco falsable para inteligencia como función del riesgo acumulado. Zenodo. https://doi.org/10.5281/zenodo.17702378
  • Rivera-García, J. M. (2026). Constitutional Architecture of Sovereign Containment for Future AI. Zenodo. https://doi.org/10.5281/zenodo.19471413
  • Krakovna, V., Langlois, E., Dafoe, A., et al. (2020). Specification Gaming: The Flip Side of AI Ingenuity. DeepMind Blog.
  • Perez, E., Ringer, S., Lukošiūtė, K., et al. (2022). Red Teaming Language Models with Language Models. arXiv:2202.03286.
  • European Union. (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act).
  • RTCA. (2011). DO-178C: Software Considerations in Airborne Systems and Equipment Certification.
  • IEC. (2010). IEC 61513: Nuclear power plants – Instrumentation and control for systems important to safety.
  • IEC. (2006). IEC 62304: Medical device software – Software life cycle processes.