NVIDIA and telecom partners are building autonomous network platforms using synthetic data, specialized AI models, secur
NVIDIA and its telecom partners are demonstrating the building blocks of autonomous networks at TM Forum's DTW Ignite 2026 in Copenhagen, showing how operators can move beyond task-based automation toward truly autonomous systems where AI agents proactively detect problems and coordinate changes across network, IT and business systems.
Telecom operators have historically achieved significant returns from generative AI automating network management, customer care and back-office operations, but most automation has been task-based, with humans manually correlating insights and directing next steps. The industry is now pushing toward genuine autonomy, supported by synthetic data, telecom-domain models, secure agent runtimes and simulations that form a complete secure autonomy platform where agents understand operator intent, act safely across business and network domains and keep humans in control of policy.
The foundation of autonomous networks is reasoning models specialized in telecom domains, requiring fine-tuning on high-quality datasets. However, 54 percent of operators cite data-related issues as their biggest barrier, with the most valuable network and customer data too sensitive to use directly. Synthetic data is enabling operators to safely increase training data volume and diversity while protecting sensitive information. SoftBank Corp. is using NVIDIA NeMo Safe Synthesizer and NVIDIA NeMo Anonymizer to generate privacy-preserving synthetic datasets that reflect the structure and distribution of real network performance and configuration data, using these datasets to fine-tune its large telecom model and build specialized network agents.
As operators pursue end-to-end autonomy, they need long-running AI agents that can stick with complex jobs from start to finish, operating under strict service-level agreements, change-management policies and regulatory constraints. NVIDIA NemoClaw blueprints and the NVIDIA OpenShell secure runtime provide these agents with policy-based guardrails and sandboxed access to telecom systems, allowing operators to safely expand agent roles while keeping behavior predictable, auditable and governed.
Multiple partners are piloting these technologies. AdaptKey is collaborating with operators on security-hardened, long-running agents for self-healing 5G network operations, with agents detecting security and connectivity issues and submitting scoped remediation requests into AdaptKey's KeySmith platform. Amdocs is showcasing NemoClaw and OpenShell for proactive customer-care agents in roaming assistance scenarios and for autonomous data-science agents that analyze customer accounts and assess migration eligibility. NTT DATA is using NVIDIA Nemotron open models with NemoClaw to build long-running agents for proactive detection of network degradation. ServiceNow is bringing Project Arc to telecom for autonomous network operations center agents that handle incident response, orchestrating the full lifecycle from initial alerts to assigned work orders while staying contained, auditable and within policy through NVIDIA OpenShell and ServiceNow AI Control Tower. Tata Consultancy Services is building a multi-fidelity AI sensor architecture using NemoClaw to help operators spot and resolve network issues faster.
Simulation is becoming integral to decision support as AI agents take on more responsibility. Forsk has integrated an AI-based radio propagation model into its Naos RAN planning platform, achieving ray-tracing-level accuracy up to 200 times faster than CPU-only baselines on NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, creating a RAN digital twin that lets operators safely optimize networks in near real time. VIAVI Solutions is accelerating its TeraVM AI RAN Scenario Generator by moving large-scale RAN simulations from CPUs to NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, achieving order-of-magnitude improvements in simulation throughput and enabling operators to run high-fidelity scenarios at real deployment scale. VIAVI has also released an IP Network Configuration Blueprint extending validation into IP and transport network domains. KDDI and KDDI Research are collaborating with NVIDIA, Keysight and Samsung Research America to build a high-fidelity RAN digital twin using NVIDIA Aerial Omniverse Digital Twin and Keysight's digital-twin-ready emulation tools, where multiple autonomous agents can safely simulate and validate RAN scenarios ranging from area-optimization strategies to future radio conditions and new AI air-interface functions.