India is recasting its IndiaAI Mission for telecom and digital infrastructure

India’s overhaul of the IndiaAI Mission is increasingly being framed through a hard telecom and digital infrastructure lens, with policymakers treating AI as a core network capability rather than just an application-layer technology. The review is being driven by concerns over dependence on foreign frontier models for AI-native networks, 6G R&D, and critical digital infrastructure management, especially after US export controls cut off foreign access to Anthropic’s latest Fable 5 and Mythos 5 systems.

AI at the core of future networks
DoT and allied agencies now explicitly position AI as a foundational element of next-generation network architecture, with 6G vision documents talking of “AI-native networks” that are largely self-optimising, self-healing, and heavily automated end to end. This shifts AI from being a support tool for analytics to becoming an embedded control layer for radio, transport, and core networks, including traffic engineering, spectrum management, and energy optimisation.

India’s 6G blueprint envisages “networks of networks” spanning terrestrial and non-terrestrial systems, with AI and machine learning used to manage complex, analytically intractable conditions and to offer AI-as-a-Service from a distributed cloud fabric. In practice, this means that whoever controls the models and compute underlying these AI-native networks effectively controls key levers of network performance, resilience, and even security posture.

Strategic alarm after Anthropic curbs
The US directive forcing Anthropic to suspend access to its Fable 5 and Mythos 5 models for all foreign nationals has acted as a wake-up call for countries that had quietly assumed persistent access to frontier cloud AI. For India’s telecom and digital infra planners, this has underscored the risk of building AI-native networks whose most advanced optimisation, planning, or security functions sit atop foreign-controlled models that can be switched off by external policy decisions.

Given ongoing work with the ITU on AI-native networks and India’s visible push to influence global 6G and satellite communications policy, the possibility of sudden model withdrawal is now seen as a direct strategic vulnerability rather than just a procurement inconvenience. This is particularly sensitive where AI is expected to drive self-healing operations, anomaly detection, and cyber defence for critical backbone networks and government communications.

Reorienting IndiaAI toward telecom infra
Against this backdrop, the IndiaAI Mission’s infrastructure components, particularly compute, data platforms, and cloud access, are being reassessed with an eye on sovereign telecom capability. The Mission already envisages large-scale GPU infrastructure and cloud-based AI services; the review is likely to push for explicit carve-outs to support telco-grade AI workloads, RAN optimisation models, and secure core-network intelligence that are fully controllable from within India.

Policy discussions are focusing on:

• Prioritising access to national AI compute for telecom operators, network OEMs, and domestic 6G R&D programmes, not just generic AI startups.
• Building indigenous foundational models tuned to Indian network conditions, traffic patterns, and spectrum usage, including rural and satellite backhaul scenarios.
• Ensuring that critical orchestration and policy engines for AI-native networks can run on Indian clouds or in operator data centres, decoupled from foreign export-control risk.

Sovereign AI for 6G and Open RAN
Government messaging around 6G already emphasises sovereign capability across chip design, secure core, satellite backhaul, and Open RAN, with AI-native networks cited as a central pillar. The IndiaAI rethink dovetails with this by treating AI models and training infrastructure as strategic assets in the same category as spectrum, silicon, and encryption technologies for future networks.

For Open RAN in particular, where intelligence is disaggregated into virtualised and cloudified components, control over AI-based RAN optimisation models and inference platforms becomes critical to avoid lock-in to foreign vendors or hyperscale clouds. Sovereign AI capabilities would allow Indian vendors and operators to train, fine-tune, and deploy their own models for beamforming, interference management, and energy control, using domestic data while retaining full control over updates and guardrails.

Implications for operators and digital infra
For operators, an overhauled IndiaAI framework aligned with telecom and digital infra priorities could translate into subsidised or priority access to national AI compute, shared model libraries for network functions, and standardised APIs for integrating AI-as-a-Service into OSS/BSS and NOC environments. At the infra layer, national AI clusters and edge compute nodes could be co-located with major fibre routes, data centres, and IXPs, creating a distributed AI fabric optimised for low-latency network control and content delivery.

The broader strategic goal is clear: ensure that as networks transition to AI-native architectures and 6G-era “network-of-networks” models, India’s telecom ecosystem is not structurally dependent on foreign-controlled AI platforms that can be restricted unilaterally. How far the IndiaAI Mission’s redesign goes in ring-fencing compute, models, and standards for telecom will likely determine the depth of India’s real sovereignty in future digital infrastructure.

CT Bureau