EUROCOMPLY: Automated regulatory compliance validation for AI/ML systems in telecommunications via agentic AI

Ameur, Mazène
ETSI AI and Data Conference 2026, 9-11 February 2026, Sophia Antipolis, France


This demonstration showcases EUROCOMPLY, an advanced Agentic AI–driven compliance validation framework purpose-built to automate regulatory conformance assessment across AI and ML pipelines in telecommunications environments. As AI-enabled network functions and services proliferate, telecom operators face increasing complexity in ensuring adherence to European regulatory mandates, including the EU AI Act, GDPR, and 3GPP AI/ML governance frameworks.
EUROCOMPLY addresses this challenge through a multi-agent compliance auditing architecture that integrates natively within MLOps ecosystems. The framework employs Large Language Model–based reasoning agents endowed with self-reflective, context-aware analytical capabilities to systematically examine datasets, training workflows, and deployment configurations. These agents operate collaboratively to detect compliance deviations, quantify associated risks, and generate evidence-based remediation recommendations mapped to both legal and telecommunication-specific regulatory standards.
By combining regulatory intelligence, AI reasoning, and engineering automation, EUROCOMPLY demonstrates how Agentic AI can act as a compliance co-pilot for telecom operators. The system not only enhances auditability and accountability in AI lifecycle management but also accelerates trustworthy AI adoption, reducing compliance overhead while mitigating legal and operational risks across complex, data-driven network ecosystems.


Type:
Poster / Demo
City:
Sophia Antipolis
Date:
2026-02-09
Department:
Systèmes de Communication
Eurecom Ref:
8595
Copyright:
Copyright ETSI. Personal use of this material is permitted. The definitive version of this paper was published in ETSI AI and Data Conference 2026, 9-11 February 2026, Sophia Antipolis, France
 and is available at :

PERMALINK : https://www.eurecom.fr/publication/8595