Sarvam AI’s Defence Deal Is Really About Who Controls India’s Military Data

Wednesday, October 28, 2026 –  India’s push for artificial intelligence self-reliance is entering a strategically important new phase. A five-year agreement between homegrown AI company Sarvam AI and Headquarters Integrated Defence Staff (HQ IDS) aims to develop AI applications for the Indian Army, Navy and Air Force. But beyond the technology announcement lies a bigger question: who will control the data, computing infrastructure and AI systems that support India’s military operations?

Announced on October 26, 2026, the agreement establishes a framework for developing AI applications across India’s three armed services. The work will cover operational support, training, knowledge management and decision-making, with human oversight retained over operational decisions.

The projects will be undertaken by Chanakya, Sarvam’s dedicated national-security and defence division, using the company’s AI technology stack and sovereign infrastructure.

The financial value of the agreement, the individual projects and the deployment schedule have not been publicly disclosed.

For India, the significance extends beyond one startup or one defence contract. The partnership provides a test of whether the country can build and operate sensitive military AI systems using infrastructure and capabilities it can control, while reducing dependence on external technology providers.

Sarvam AI Defence Deal: What the Five-Year Agreement Covers

The agreement between Sarvam AI and HQ IDS creates an umbrella framework for multiple AI projects across the Indian Army, Navy and Air Force, including joint and integrated defence organisations.

Under the arrangement, Sarvam will design, develop and deploy AI applications tailored to the operational, training and decision-support requirements of the services. The company will also provide technical support after deployment.

The proposed areas of work include:

• Decision support: Helping military personnel analyse information and assess relevant operational scenarios.
• Data fusion: Bringing together information from different sources to support a more comprehensive understanding of a situation.
• Target acquisition support: Applying AI to assist processes involved in identifying and assessing potential targets.
• Training and knowledge management: Using AI applications to organise information and support learning across defence organisations.
• Shared AI infrastructure: Providing the three services with access to common foundation models, agentic infrastructure, geospatial systems and perception technologies.

One of the stated objectives is to build an AI layer that can connect different elements of India’s defence capabilities. This could enable lessons from projects developed for one service to inform work across the others.

However, the agreement should not be mistaken for a disclosed procurement of a fully operational autonomous weapons system. The announced scope focuses on AI applications and supporting capabilities, with human oversight retained over operational decisions.

Why Military Data Sovereignty Matters

Military data is not simply another category of enterprise information. It can contain details about operational procedures, intelligence assessments, equipment performance, logistics, geographical conditions and the way armed forces respond to different situations.

When this information is processed by AI systems, it may also influence how models are evaluated, adapted and improved.

That makes control over the entire technology environment important. A country may develop its own AI model but still rely on external providers for cloud hosting, computing infrastructure, software components or essential technical services.

The key question is therefore not just where an AI model was created. It is whether the organisation deploying it can control access to its data, manage the infrastructure on which it runs, enforce security requirements and continue operating if an external supplier becomes unavailable.

For India’s defence establishment, stronger domestic control could offer several potential advantages:

Greater control over sensitive information: Defence organisations can establish rules governing where data is stored, who can access it and how it may be processed.

Reduced exposure to external dependencies: Domestic deployment can reduce reliance on foreign-hosted services, although hardware and software dependencies may remain.

More control over system development: Indian institutions and technology partners can adapt AI applications to local requirements, languages and operational contexts.

Continuity of critical services: Infrastructure under domestic control can provide greater resilience against certain external disruptions, depending on how the systems are designed and operated.

These benefits are not automatic. Sovereignty also depends on cybersecurity, supply-chain security, access controls, maintenance arrangements and the ability to audit systems throughout their operational life.

Chanakya: Sarvam’s Dedicated Defence AI Division

The defence projects will be handled by Chanakya, Sarvam’s dedicated division for national-security and defence applications.

Unlike conventional cloud-based AI services, defence systems may need to operate inside tightly controlled environments that restrict or entirely prevent connections to the public internet.

One approach is an air-gapped deployment, in which a system is isolated from external networks. Depending on the implementation, such an environment can help limit unauthorised data transfers and reduce exposure to some network-based threats.

Sarvam has described Chanakya as focused on on-premises and air-gapped deployments. The precise security architecture, certification status and implementation arrangements for the new projects have not been publicly detailed.

Chanakya will use Sarvam’s AI technology stack, including foundation models such as Sarvam 105B. The company has previously described Sarvam 105B as a model developed in India and released as open source.

The defence partnership also envisages common access to AI models, agentic infrastructure and geospatial and perception systems across the services.

A shared foundation could help reduce fragmentation between projects. However, effective integration will depend on data standards, compatibility with existing systems, security approvals and the ability to adapt the technology to each service’s requirements.

India’s AI Ambitions Face a Compute Challenge

Developing domestic AI models is only one part of building a sovereign AI ecosystem. Training and running advanced models also require substantial computing capacity, specialised chips, energy, data centres and supporting software.

India has been expanding access to AI computing through the IndiaAI Mission. Government information published in 2026 reported that more than 38,000 GPUs had been empanelled through AI service providers to supply shared computing infrastructure.

That capacity supports a broader national effort to develop AI models and applications within India. Sarvam’s foundation-model work has also received support through the IndiaAI ecosystem.

Nevertheless, domestic access to computing does not necessarily mean that every part of the hardware supply chain is indigenous. Advanced accelerators, semiconductor manufacturing and certain software components remain areas where international suppliers play an important role.

This distinction matters when assessing claims of technological sovereignty.

A model can be developed by an Indian company, trained using computing capacity available in India and deployed within a domestically controlled environment, while still depending on foreign-designed hardware.

That can represent meaningful progress in data control and operational independence without amounting to complete hardware self-sufficiency.

For the defence sector, the practical objective is to understand and manage these dependencies rather than assume that a locally developed model automatically removes them.

Why a Shared AI Framework Across the Three Services Matters

The Army, Navy and Air Force have different operational requirements, equipment and information systems. Their AI applications may therefore need to be tailored to distinct environments.

At the same time, some underlying capabilities—such as language processing, information retrieval, document analysis and certain forms of data integration—may be useful across multiple defence organisations.

The Sarvam-HQ IDS framework seeks to provide access to common AI foundations while supporting projects across the three services.

If implemented effectively, a shared technology layer could help organisations reuse validated capabilities, reduce duplicated development and make it easier to transfer lessons between projects.

It could also create new integration challenges. Defence organisations must manage different data classifications, legacy systems, operational procedures and security requirements. A common AI foundation does not eliminate the need for service-specific controls or rigorous testing.

The agreement’s long-term value will depend on whether it can make collaboration easier without weakening security or operational accountability.

The International Debate Over Military AI and Vendor Dependence

India is not alone in examining the relationship between military AI, private technology companies and national control.

Defence organisations worldwide are increasingly using commercial technology providers for data analysis, software platforms and AI-enabled decision support. These partnerships can accelerate development, but they may also create long-term dependencies.

If one supplier controls critical software, infrastructure or integration layers, changing providers can become expensive and technically difficult. A customer may also have limited visibility into how certain systems work or how their underlying components are maintained.

These concerns have featured in international debates over military AI platforms, including discussions about US defence technology providers and NATO-related projects.

For India, the lesson is not that private companies should be excluded from defence technology. Domestic startups can contribute specialist expertise and develop products that public institutions may struggle to build alone.

The important question is how the relationship is structured.

Defence customers need clear arrangements covering data ownership, system access, security audits, intellectual property, updates, incident response and the ability to maintain or replace critical components.

A sovereign AI strategy is strongest when domestic technology providers can compete and innovate without becoming unavoidable single points of dependence.

What the Agreement Does Not Yet Reveal

Although the five-year framework establishes an important direction, several details remain unknown.

Neither Sarvam nor HQ IDS has publicly disclosed the agreement’s financial value, a complete list of initial projects or a detailed timetable for deployment.

The public announcement also does not provide a comprehensive description of the data-governance arrangements, computing hardware, security certification process or the precise architecture of the proposed AI systems.

These details will matter when assessing how much operational independence the partnership can deliver.

It will also be important to distinguish between an announced framework and proven results. Developing an AI application is only the first step; the technology must then be tested, integrated, secured and evaluated under the conditions in which it is intended to operate.

The five-year period provides a structure for collaboration, but the pace of implementation and the scale of eventual deployment remain to be established.

The Bigger Picture: India’s Sovereign AI Strategy

The Sarvam AI defence agreement reflects a broader shift in India’s technology priorities. The country is seeking to develop domestic AI capabilities that can support important national functions while strengthening control over sensitive information and infrastructure.

For the defence sector, this ambition involves more than creating an Indian alternative to a foreign AI model. It requires a complete environment covering models, data governance, computing, cybersecurity, integration and long-term technical support.

The partnership could help build that environment by bringing a domestic AI developer into projects spanning the Army, Navy and Air Force.

Its success, however, will depend on execution. Security arrangements must work in practice, models must perform reliably, and defence organisations must retain meaningful oversight over how AI recommendations are used.

The key measure of progress will not simply be whether the technology is developed in India. It will be whether India’s defence institutions can operate, secure, evaluate and sustain these systems with the level of independence their missions require.

Ultimately, the strategic question behind India’s military AI push is straightforward: who controls the data, the infrastructure and the systems on which national security increasingly depends?

India Opens 111 New GCCs in 2026, Surpassing 2025’s Full-Year Total