Abstract

AI Is Beginning to Control Mobile Networks: Who Proves a Change Works Before It Reaches Millions of Subscribers?

Technical Subtitle: From Qualcomm Dragonwing Agentic RAN and Ericsson Intelligent Automation Platform to Nokia MantaRay SMO — How Receipt-Gated Execution, Authenticated Live-Network Evidence, Independent Validation and Protected Rollout Authority Could Strengthen 5G-Advanced, Open RAN, Network Slicing and Future 6G Infrastructure.

ABSTRACT

Mobile telecommunications networks are entering a new phase of artificial intelligence-driven automation. AI agents, intelligent orchestration platforms, and autonomous network-management applications are increasingly capable of proposing and executing consequential changes across radio access networks, core-network functions, network slices, traffic-management systems, and service-assurance infrastructure.

Technology leaders including Qualcomm, Ericsson and Nokia are advancing AI-enabled RAN optimization, agentic network management, closed-loop automation, programmable network infrastructure, and increasingly autonomous telecommunications operations.

These developments create an important technical distinction between authorizing a network change, receiving confirmation that a management command was accepted, and independently verifying that the change actually produced the approved operational outcome in the live network.

An AI-generated optimization plan may be valid according to simulation or network policy while its actual implementation produces an unexpected effect on coverage, throughput, latency, mobility, energy consumption, or service availability.

A network controller may acknowledge a configuration change even though the intended operational state was not established. Streaming telemetry may indicate apparent success without proving that the reported observation corresponds to the exact authorized action, resource, configuration version, or deployment scope.

When an autonomous operation is expanded across thousands of network elements, even a narrowly misconfigured action can propagate into consequential service degradation, capacity loss, or disruption affecting enterprise and critical infrastructure connectivity.

Conventional authorization, orchestration, observability, digital-twin simulation, and closed-loop optimization mechanisms remain essential. However, they do not, merely by existing, establish a protected dependency between independently verified evidence of an earlier bounded real-network effect and the authority to release a broader consequential network change.

This technical study examines the proposed Receipt-Gated Interim Effectuation Validation (IEV) architecture and its prospective integration into three important telecommunications technology environments represented by Qualcomm, Ericsson and Nokia.

The proposed mechanism separates an AI-generated candidate network action from authority to apply that action across its full intended scope.

A protected controller first permits a small, real, strictly bounded effect, such as a controlled configuration adjustment at one site, a limited traffic-steering operation, a restricted core-network policy change, or the activation of a network slice for a selected service cohort.

The affected network element, management system, or protected independent observer then generates authenticated Effect-Confirmation Evidence, potentially represented by an Effect-Confirmation Receipt (ECR).

An independent validator checks this evidence against the approved operation, target network elements, configuration parameters, observed service state, permitted scope, freshness, policy conditions, and relevant operational safety thresholds.

Only after the required validation succeeds does a protected enforcement component authorize broader deployment.

The central technical distinction is not another network-policy approval or ordinary canary deployment. It is making independently validated evidence of an earlier real-network effect a mandatory technical condition for releasing subsequent network-execution authority.

The study examines three distinct implementation pathways.

Qualcomm — Agentic RAN Optimization and Controlled Network-Wide Changes: A proposed protected management service operates at a supported automation boundary associated with Dragonwing RAN Automation Suite and Agentic RAN Management Service. AI-generated optimization or energy-management actions are first applied within a limited live scope. Authenticated network-state and service-quality evidence must satisfy the approved conditions before wider rollout is authorized.

Ericsson — Cross-Domain RAN and Core Automation: A proposed protected action broker operates at a supported execution boundary associated with Ericsson Intelligent Automation Platform, rApps, cApps, and network-management interfaces. Limited traffic-steering or core-network policy changes are validated against authenticated effect evidence before the system permits broader RAN or core-network consequences.

Nokia — Network Slice Lifecycle and Autonomous Service Assurance: A proposed protected execution layer operates alongside supported MantaRay and RAN Slice Controller interfaces. A slice activation, resource adjustment, or autonomous remediation is initially restricted to a selected slice, cell, device population, or service cohort. Independently validated slice-state and service-performance evidence governs subsequent expansion.

The proposed architecture is designed to complement existing network optimization algorithms, RAN and core functions, O-RAN management interfaces, radio hardware, security systems, digital twins, and telecommunications standards.

Importantly, the mechanism is intended primarily for consequential network-management, orchestration, configuration, and rollout timescales.

It is not proposed as an additional receipt-validation step for every radio scheduling decision, physical-layer operation, fronthaul transaction, or user-plane packet.

The study further evaluates forged telemetry, alternate management paths, receipt replay, lost acknowledgements, cross-domain consistency, performance overhead, legacy interoperability, and potential hardware-rooted enforcement where supported interfaces permit.

The principal research question is whether autonomous telecommunications systems can progress beyond predicting and approving network changes toward independently verifying limited real-network outcomes before granting authority to expand those changes across critical infrastructure.

The company-specific integrations are proposed engineering architectures and require independent implementation, measurement, and validation.

Part 6 of 8 — AI-Native RAN, Autonomous Core Networks and Protected 6G Execution

This is Part 6 of an eight-part technical disclosure series.

Part 1 introduces the foundational receipt-gated execution concept for general readers.

Parts 2–5 examine prospective applications involving frontier AI platforms, enterprise cloud systems, payment networks, accelerated computing, and telecommunications infrastructure.

Part 6 focuses on three major telecommunications technology providers — Qualcomm, Ericsson and Nokia — and examines how authenticated outcome evidence could complement their distinct approaches to AI-driven network automation.

This study concentrates on three implementation challenges:

  • Qualcomm: Verifying bounded RAN optimization or energy-management outcomes before expanding autonomous changes across additional sites.

  • Ericsson: Establishing trustworthy execution evidence across RAN and core-network automation, where an accepted management command may not demonstrate the intended end-to-end service effect.

  • Nokia: Validating live network-slice state and service-assurance requirements before permitting wider slice activation, resource expansion, or autonomous remediation.

The emphasis is on practical enforcement boundaries, multi-vendor compatibility, minimal disruption to legacy infrastructure, and avoiding additional latency in time-critical radio functions.

The Core Idea in Plain Language

Imagine an AI network-management system proposing a configuration change for 10,000 mobile network sites.

A simulation indicates that the change should improve energy efficiency or network performance.

But simulation success does not prove that the configuration will produce the intended result on every real site.

Instead of immediately applying the operation across the entire network, the proposed architecture follows four steps.

1. Execute a small real change.

The protected network controller permits the change on one approved site, network slice, limited subscriber cohort, or another strictly bounded operational scope.

2. Obtain authenticated evidence from the actual network.

A network element, protected management system, or independent observer generates evidence identifying the applied configuration, affected resources, actual observed network state, and relevant operational conditions.

3. Validate the evidence independently.

A validator outside the AI agent's control checks whether the observed outcome matches the original authorization, including destination, configuration version, scope, freshness, and required service-quality conditions.

4. Authorize broader deployment conditionally.

Only after the evidence satisfies the protected validation conditions does the enforcement component permit a wider rollout.

No valid qualifying receipt means no authorization to expand the consequential operation.

If the acknowledgement is lost, the system does not blindly repeat the operation. It checks the authoritative configuration and operational state, distinguishes failure from an indeterminate outcome, and holds the broader rollout until reconciliation requirements are satisfied.

The initial verified result is evidence of a bounded real outcome, not a guarantee that every subsequently affected site will behave identically. Later phases remain subject to their own protection and validation requirements.

Industry Alignment 1 — Qualcomm: Dragonwing Agentic RAN and Protected Network Optimization

Relevant Technology Directions: Qualcomm Dragonwing RAN Automation Suite, Agentic RAN Management Service, QDU100 Telco Server Platform, QRU100 Radio Unit Platform, AI-RAN, autonomous RAN optimization, and future 6G-native network research.

Qualcomm's RAN automation and commercial radio platforms provide a relevant environment for evaluating how bounded real-network evidence could govern the progression of AI-generated management actions.

Proposed Technical Integration

An AI-driven RAN management system proposes a network optimization or energy-saving configuration affecting multiple sites.

The proposed protected execution layer operates at an authorized management or orchestration interface, before the change can be released across the full intended network scope.

The controller initially permits the configuration on one selected site or another strictly limited operational cohort.

An authenticated network-state record or protected observation mechanism establishes evidence of the actual change and its observed effect.

The independent validator evaluates:

  • The identity of the affected site and network elements.

  • The applied configuration and approved parameter scope.

  • Relevant traffic, coverage, and service-quality measurements.

  • Operational thresholds, policy conditions, and evidence freshness.

  • Whether the observed result justifies progression to the next approved cohort.

Only when the required evidence is accepted does the protected controller authorize broader rollout.

A successful optimization recommendation or management command is not automatically treated as proof that the desired live-network improvement occurred.

Practical Enforcement Boundary

The proposed mechanism operates above time-critical radio functions, within supported management or orchestration controls.

It does not require inserting receipt checks into each modem operation, RU or DU processing cycle, physical-layer calculation, or radio scheduling interval.

Qualcomm's existing hardware and low-level RAN optimization mechanisms can therefore remain unchanged in a software-first proof of concept, subject to the availability of appropriate management interfaces.

Deeper hardware-rooted enforcement would require separate technical integration and validation.

Industry Alignment 2 — Ericsson: Independent Evidence Across RAN and Core Automation

Relevant Technology Directions: Ericsson Intelligent Automation Platform (EIAP), rApps, cApps, Ericsson Network Manager, closed-loop automation, network observability, and autonomous RAN/core operations.

Ericsson's expansion of automation across both radio access and core-network domains provides a particularly relevant environment for examining the difference between a management action being issued and its intended service consequence being independently established.

Proposed Technical Integration

A protected action broker is positioned at a supported execution boundary for selected rApp or cApp operations.

For example, an rApp may propose a traffic-steering adjustment across several cells, or a cApp may propose modifying a core-network policy affecting a service population.

The protected broker initially permits only a limited real change, such as one site, selected traffic flow, restricted session group, or approved service cohort.

The destination network systems or protected independent observers produce evidence describing the applied policy, affected resources, operational state, and resulting service conditions.

An independent validator evaluates this evidence before the controller authorizes expansion.

Streaming telemetry alone is not automatically equivalent to authenticated proof of the exact intended effect.

Telemetry may become qualifying evidence when its provenance, integrity, observation scope, time binding, and relationship to the specific authorized operation have been established.

Cross-Domain Execution Challenge

A RAN configuration may appear correctly applied while a corresponding core-network policy remains inconsistent.

Similarly, a core controller may acknowledge a policy update even though the intended service behavior is not observed at the access-network boundary.

The proposed architecture therefore supports validating evidence from more than one protected network domain when the authorized outcome depends on cross-domain consistency.

A wider change can remain blocked until the necessary RAN and core evidence satisfies the approved requirements.

The technical objective is to bind continuation authority to validated service consequences, rather than to an isolated command acknowledgement.

The architecture would supplement existing Ericsson automation and network-management capabilities without assuming replacement of rApps, cApps, or network observability infrastructure.

Industry Alignment 3 — Nokia: MantaRay, Network Slicing and Autonomous Service Assurance

Relevant Technology Directions: Nokia MantaRay SMO, MantaRay AutoPilot, MantaRay SON, RAN Slice Controller, Autonomous Networks Suite, Autonomous Networks Agent Library, and Open RAN automation.

Network slicing represents an important area for consequence-bound authorization because slice creation, expansion, modification, and remediation can affect distributed network resources and contractual service requirements.

Proposed Technical Integration

An autonomous orchestration system proposes activating or expanding a network slice to support additional devices, subscribers, industrial applications, or enterprise services.

Rather than immediately granting the complete requested resource scope, the proposed protected controller permits a limited real activation.

The initial phase may affect one cell, selected device group, specific slice subnet, or controlled service cohort.

A protected observer or supported network-management system generates authenticated evidence of the actual activated resources, slice configuration, and relevant operational state.

An independent validator compares the observed conditions with the approved slice requirements.

These requirements may include throughput, latency, reliability, resource allocation, service availability, policy conditions, and relevant service-level targets.

Only after the evidence satisfies the required conditions does a protected enforcement component authorize the next expansion phase.

Autonomous Remediation

The same general mechanism could apply to remediation actions proposed by autonomous-network agents.

For example, an agent may recommend reallocating resources to address a detected service degradation.

A protected execution boundary first authorizes a strictly limited corrective change.

The validator then determines whether the real operational state changed within the approved constraints before additional remediation authority is released.

The intended distinction is between proposing a corrective action, applying it, and independently establishing that its actual effect satisfies the authorized conditions.

Existing Nokia optimization functions, SON mechanisms, slice-management components, and autonomous-network frameworks would remain complementary rather than being replaced.

Why This Is Different From Ordinary Closed-Loop Automation

Conventional closed-loop automation commonly observes network conditions, makes a decision, applies a change, and monitors the outcome.

Canary deployment and progressive rollout are also established engineering practices.

The proposed technical distinction is narrower and more specific.

An earlier bounded real-network effect produces authenticated evidence that must be independently validated before the system can establish the protected authority needed for the next consequential phase.

The agent's initial authorization does not automatically include authority for the full rollout.

The AI application cannot simply report success and unlock the next stage.

The continuation state is maintained outside the agent's control, and the downstream enforcement boundary must reject broader execution without qualifying authorization.

This distinction requires practical testing because an application-level validator cannot prevent bypass if an alternate privileged management interface can independently apply the same change.

6G and IMT-2030: Research Relevance, Not Standards Adoption

Future 6G networks are expected to involve increasingly sophisticated intelligence, automation, distributed infrastructure, and service management.

These developments create opportunities to investigate consequence-bound authorization for AI-driven network-management operations.

However, the proposed receipt-gated mechanism is not represented as a requirement, approved feature, or adopted component of any 3GPP, O-RAN, or IMT-2030 specification.

Its potential relevance to future 6G infrastructure remains a prospective research direction.

Practical implementation must preserve existing real-time radio operation, local control loops, mobility responsiveness, and other latency-critical functions.

Additional validation should be placed only where the operational consequence and available timing budget justify the enforcement cost.

Technical Scope and Validation Programme

The study also addresses the requirements necessary to evaluate practical implementation.

Shared Security Architecture: Candidate network actions, bounded real-effect execution, authenticated receipts, independent validation, protected continuation state, and controlled rollout authority.

Evidence and Protocol Formats: Operation identity, target network element, configuration version, phase identifier, observed state, telemetry provenance, policy epoch, freshness, nonce, counter, and prior receipt binding.

Product Integration Matrix: Proposed entry points, evidence sources, supported management interfaces, continuation-state locations, and potential enforcement limitations for Qualcomm, Ericsson, and Nokia.

Adversarial Threat Testing: Fabricated telemetry, forged success acknowledgements, alternate management credentials, stale receipts, replay, configuration substitution, partial execution, and attempts to bypass the protected controller.

Indeterminate-Outcome Recovery: Reconciliation of lost acknowledgements, inconsistent network states, disconnected sites, incomplete activation, and ambiguous configuration outcomes before further rollout.

Latency and Performance: Separating management-level validation from fast radio scheduling, physical-layer processing, fronthaul behavior, and user-plane packet processing.

Legacy and Multi-Vendor Compatibility: Integrating through supported management, orchestration, SMO, RIC, network-slice, and operational interfaces without assuming replacement of installed radio hardware.

Recommended Proof-of-Concept Sequence: A single-site RAN optimization, a cross-domain RAN/core policy validation, and a limited network-slice activation followed by evidence-gated expansion.

The programme should independently measure bypass resistance, false-success rejection, evidence freshness, transaction and state reconciliation, rollout overhead, service impact, and operational feasibility.

No claim of universal non-bypassability, zero latency, guaranteed performance improvement, or complete interoperability is made without supporting experimental evidence.

Foundational Patent and Technical References

Mothership International Patent Application PCT/IB2026/055615 Filed: 4 June 2026 Publicly disclosed: 16 July 2026

Related Indian Provisional Patent Application 202631117633 Filed: 1 October 2026

Related IETF Internet-Draft

Evidence Is the Key, Not the Log — Receipt-Gated Staged Effectuation and Interim Effectuation Validation for AI Agents, Frontier AI Model Providers, and Autonomous Systems

Internet-Draft Identifier: draft-das-receipt-gated-iev

Document dated 1 October 2026; Datatracker records an update on 8 October 2026.

Read the IETF Internet-Draft

The document is an individual experimental Internet-Draft and does not represent IETF endorsement, consensus, or an adopted standard. The original public availability date and patent-priority implications require independent assessment.

Patent-Pending Technical Concept — DAS Protocols

Non-Endorsement, Independent Verification and AI-Assistance Notice

Company and product names appear exclusively to illustrate prospective technical integration points and potential industry relevance.

Nothing in this publication states or implies that Qualcomm, Ericsson, or Nokia has implemented, adopted, licensed, partnered on, endorsed, or independently validated the proposed invention.

The publication does not allege copying, patent infringement, security deficiencies, or technical non-compliance by any named company.

The company-specific alignment analysis was prepared with AI assistance and may contain assumptions, inaccuracies, or product descriptions requiring independent verification.

Readers should independently cross-check the proposed mappings against the original patent disclosures, current official vendor documentation, supported product interfaces, applicable telecommunications standards, and reproducible experimental results before drawing technical, commercial, or legal conclusions.

Access to the Foundational PCT Filing

The foundational patent materials may be located through WIPO PATENTSCOPE using:

Inventor: Sangam Das PCT Application Number: PCT/IB2026/055615 Alternative Search: pctib202655615

Additional technical publication:

TD Commons — Foundational PCT Reference

Strategic Collaboration and Licensing

The inventor is open to strategic technical collaboration, technology licensing discussions, and acquihire opportunities involving AI-native telecommunications, 5G-Advanced and future 6G infrastructure, Open RAN, autonomous network management, semiconductor platforms, and critical infrastructure security.

Publication License: Creative Commons Attribution–NonCommercial 4.0 International (CC BY-NC 4.0).

This publication licence does not grant patent or commercial practice rights. Commercial implementation of patent-protected subject matter requires a separate written licence from the relevant rights holder.

Keywords and Technical Search Index

Core Architecture: Receipt-Gated Execution; Receipt-Gated IEV; Interim Effectuation Validation; Effect-Confirmation Receipt; Staged Execution; Bounded Real-Effect Validation; Canary Release; Progressive Rollout; Conditional Commit; Signed Receipt; Authenticated Outcome Evidence; Independent Validation; Protected Execution-Finality; Protected Continuation Authority; Cryptographic Release Authorization.

Telecommunications Security: AI-RAN; Agentic RAN; Autonomous Networks; AI-Native Networks; Network Slicing; O-RAN; SMO; Non-RT RIC; rApps; cApps; Closed-Loop Automation; Network Orchestration; Live Network-State Verification; Service Assurance; Network Configuration Finality; Zero Trust; Policy Enforcement Point; TOCTOU Prevention.

Qualcomm Technologies: Qualcomm; Qualcomm Dragonwing; Dragonwing RAN Automation Suite; Agentic RAN Management Service; QDU100; QRU100; AI-Driven RAN Optimization; RAN Energy Management; Autonomous RAN Supervision; AI-RAN; Open RAN; Radio Unit; Distributed Unit; RAN Digital Twin; RAN Automation.

Ericsson Technologies: Ericsson; Ericsson Intelligent Automation Platform; EIAP; Ericsson Network Manager; Ericsson Stream Processing and Enrichment; ESPE; rApps; cApps; RAN-Core Automation; Intelligent RAN; Closed-Loop Network Automation; Core Policy Validation; Cross-Domain Network Assurance; Autonomous Core Networks.

Nokia Technologies: Nokia; Nokia MantaRay SMO; MantaRay AutoPilot; MantaRay SON; Nokia RAN Slice Controller; Autonomous Networks Suite; Autonomous Networks Agent Library; Non-Real-Time RIC; Network Slice Lifecycle; Slice SLA Assurance; Dynamic Resource Orchestration; Autonomous Remediation; Multi-Vendor RAN Management.

Future Networks and Research: 5G-Advanced; 5G-A; 6G; IMT-2030; 3GPP; O-RAN Alliance; AI-Native 6G; Network Digital Twin; Telecom Foundation Models; Autonomous Network Operations; Edge Intelligence; Critical Telecommunications Infrastructure.

Patent and Standards References: DAS Protocols; Patent Pending; PCT/IB2026/055615; Indian Provisional Application 202631117633; Evidence Is the Key, Not the Log; draft-das-receipt-gated-iev.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Share

COinS