Indian Firm Neuralix Develops Sovereign AI Foundation Model for Indian Army Networks

India’s effort to build sovereign artificial-intelligence capabilities for its armed forces has moved into a new phase, with Neuralix AI Private Limited announcing the development of a military-specific foundation model designed to operate entirely within the Indian Army’s secure digital environment.

The company said on September 2, 2026 that the AI model has been built specifically around military requirements rather than by simply adapting a commercially available general-purpose language model. It forms part of a broader indigenous defence-AI ecosystem being developed under the Ministry of Defence’s Innovations for Defence Excellence, or iDEX, ADITI 2.0 programme.

The programme itself has official backing. The Ministry of Defence’s ADITI 2.0 challenge documents identify the Indian Army’s Directorate General of Information Systems, or DGIS, as the organisation seeking an “AI as a Service along with Infra Setup” capability. The requirement calls for a unified AI platform capable of integrating multiple artificial-intelligence services with Army applications and supporting decision-making, automated data analysis and operational efficiency.

That requirement eventually led to Project EKAM, which was formally launched by Defence Minister Rajnath Singh at the Chanakya Defence Dialogue on November 28, 2025.

In its official announcement at the time, the Ministry of Defence described EKAM as a strategic initiative intended to develop indigenous artificial-intelligence solutions for the Armed Forces. The ministry highlighted the importance of retaining control over data security while developing AI systems tailored to military terminology, operational doctrines and defence-specific datasets.

According to the government’s description, EKAM provides a framework and platform capable of deploying and hosting both open-source and indigenous AI models on advanced AI applications. The objective is to give the Armed Forces a controlled environment in which different AI capabilities can be integrated without having to depend entirely on external commercial platforms.

Neuralix says the latest development goes a step further by adding a military-specific foundation model to this sovereign infrastructure.

According to the company, the model was built from the ground up around defence requirements and is intended to understand military terminology, operational concepts, doctrine, strategy and complex Army documents. It has also been designed to generate structured information and decision-support outputs for authorised military users.

The significance of this approach lies largely in where the system operates.

Unlike many commercial generative-AI products that rely on public cloud infrastructure, internet-connected model repositories and external application-programming interfaces, Neuralix says its military AI stack has been designed to operate inside an air-gapped Indian Army computing environment.

An air-gapped network is deliberately isolated from public networks and the internet. Such an architecture is common in highly sensitive defence and national-security systems because it reduces the possibility that classified or operational data could inadvertently leave the controlled network.

Neuralix says both the foundation model and EKAM can operate without depending on external cloud services or public internet connectivity. Under this architecture, sensitive Army information can remain within Army-controlled infrastructure rather than being transferred to outside AI platforms for processing.

This distinction is particularly important as artificial intelligence begins to move from administrative applications towards military decision-support systems.

Defence organisations can potentially use generative AI to analyse large collections of documents, summarise intelligence, retrieve technical information, translate material, support logistics planning and assist commanders in processing complex information. However, these applications also involve information that may be operationally sensitive.

A sovereign AI system therefore involves more than simply developing an Indian software interface. Genuine technological sovereignty requires control over the model, the data and the computing infrastructure on which the system operates.

Neuralix describes its architecture in similar terms. The company says control of the model architecture and weights represents one layer of sovereignty, while keeping military information inside Army networks provides data sovereignty. Running the entire system on controlled defence infrastructure provides a third layer by reducing dependence on foreign cloud platforms and external AI services.

The company has also developed EKAM as a unified military generative-AI platform, capable of hosting different AI models and specialised services within the Army’s digital environment.

Rather than functioning as a single chatbot or application, the platform is intended to act as an underlying AI layer onto which individual military applications can be added. These could eventually include specialised language models, military AI agents, document-analysis tools and other decision-support systems developed for specific Army requirements.

The Ministry of Defence’s original ADITI challenge points in the same direction. The Army requirement called for a platform that could integrate AI services with multiple applications, generate actionable insights, automate selected processes and serve as a core element of a wider decision-support ecosystem.

Neuralix says it was selected as the technology developer through the iDEX competitive process, with DGIS functioning as the steering and user agency and the Defence Innovation Organisation acting as the contracting agency.

The company’s role in the programme is also supported by earlier public information associated with the iDEX ecosystem. Neuralix was identified as the winner of the ADITI 2.0 Indian Army challenge for AI as a Service along with infrastructure setup, while Project EKAM was subsequently officially unveiled by the Defence Minister.

The latest foundation model is therefore best understood as part of a longer development programme rather than an isolated announcement.

Neuralix says its engineers developed and refined the system while working within the secure environment provided by DGIS. This approach was intended to allow the development team to work around military requirements without removing sensitive Army data from the controlled infrastructure.

The company also says it completed development of the EKAM platform and the military foundation model over a period of slightly more than a year.Armanetics

Some of the more technical details released on September 2, including the model’s internal architecture, development methodology and specific military knowledge capabilities, currently originate from Neuralix’s own disclosure and have not been independently detailed by the Ministry of Defence in a new public release. These claims should therefore be treated as company-provided technical information unless additional Army or Ministry documentation becomes available.

Even with that qualification, the broader programme has considerable strategic importance.

Modern foundation models increasingly form the underlying intelligence layer for applications ranging from conversational systems to autonomous agents and advanced information-analysis tools. If a military depends entirely on externally hosted foundation models, it can create concerns over data control, long-term access, software dependencies and the ability to modify systems independently.

Developing an indigenous model inside a controlled military computing environment attempts to address many of these concerns.

It could also provide an architecture on which more specialised Indian defence-AI capabilities can eventually be developed. Instead of creating each AI application independently, the Armed Forces could potentially build domain-specific systems on top of a common sovereign AI infrastructure.

Future applications could theoretically include logistics support, technical-document analysis, multilingual processing, maintenance intelligence, operational planning assistance and knowledge-management tools. Any extension into operational command or weapons-related systems, however, would require separate testing, evaluation, security certification and military approval.

The development also illustrates the expanding role of Indian startups and smaller technology companies in areas that were once dominated almost entirely by large defence organisations.

Through programmes such as iDEX and ADITI, the Ministry of Defence has increasingly attempted to connect startups with specific technological requirements identified by the Armed Forces. Rather than developing technology first and searching for a customer later, companies are often working against defined military problem statements.

In Neuralix’s case, the original Indian Army requirement was explicit: create a unified AI-as-a-Service platform capable of supporting secure AI-powered decision-making inside the Army’s own digital ecosystem.

The emergence of EKAM and the subsequent development of a military-specific foundation model indicate how that requirement is now evolving into a broader sovereign defence-AI architecture.

For India, the strategic value may ultimately lie less in producing another large language model than in demonstrating that advanced AI systems can be developed, hosted and controlled entirely within national defence infrastructure.

As artificial intelligence becomes increasingly important to future military operations, maintaining control over the underlying models, training environment, data and computing infrastructure could become as important as the individual applications built on top of them.

Neuralix’s latest announcement therefore represents an important development in India’s wider push towards indigenous defence technology. The officially confirmed EKAM programme has already established the framework for sovereign AI services within the Indian Army, while the company’s new military foundation model seeks to provide the specialised intelligence layer that could support a wider generation of secure Indian defence-AI applications.

---

References

Press Information Bureau — Ministry of Defence
India has become a voice of balance & responsibility in the current global environment; Raksha Mantri at Chanakya Defence Dialogue
Published: 28 November 2025

The official release confirms the launch of EKAM: AI as a Service — iDEX ADITI 2.0 and describes it as an indigenous AI initiative for the Armed Forces designed to provide control over data security and support military-specific terminology, doctrines and datasets.

Innovations for Defence Excellence — ADITI 2.0, Ministry of Defence
Problem Statement 4: AI as a Service along with Infra Setup

The official challenge document identifies the Indian Army as the user organisation and describes DGIS’s requirement for a unified AI platform capable of integrating multiple AI services and supporting military decision-making and automated analysis.

Neuralix AI Private Limited — Company Announcement
Neuralix Deploys Sovereign AI Foundation Model Built Ground-Up for Military Use
Published: 2 September 2026

---