The Case for a Horizontal Federated AI operating System for Telcos

TL;DR

This paper advocates for a horizontal federated AI operating system tailored for telcos, aiming to unify AI efforts, ensure data privacy, and enable scalable, compliant AI deployment across telecom networks.

Contribution

It introduces a novel federated AI operating system architecture for telcos, supporting interoperability, compliance, and multi-vendor extensibility, which is a significant shift from vertical vendor platforms.

Findings

Proposes a federated AI platform architecture for telcos.

Supports federated training and compliance with industry standards.

Enables scalable, multi-vendor AI deployment in telecom networks.

Abstract

As artificial intelligence capabilities rapidly advance, Telco operators face a growing need to unify fragmented AI efforts across customer experience, network operations, and service orchestration. This paper proposes the design and deployment of a horizontal federated AI operating system tailored for the telecommunications domain. Unlike vertical vendor-driven platforms, this system acts as a common execution and coordination layer, enabling Telcos to deploy AI agents at scale while preserving data locality, regulatory compliance, and architectural heterogeneity. We argue that such an operating system must expose tightly scoped abstractions for telemetry ingestion, agent execution, and model lifecycle management. It should support federated training across sovereign operators, offer integration hooks into existing OSS and BSS systems, and comply with TM Forum and O-RAN standards. Importantly, the platform must be governed through a neutral foundation model to ensure portability, compatibility, and multi-vendor extensibility. This architecture offers a path to break the current silos, unlock ecosystem-level intelligence, and provide a foundation for agent-based automation across the Telco stack. The case for this horizontal layer is not only technical but structural, redefining how intelligence is deployed and composed in a distributed network environment.