In March 2026, a prominent chatbot in the US got blacklisted by the Pentagon for refusing to loosen safeguards on military use of its AI model. The blacklisted chatbot soon became No. 1 in US app downloads as users flocked to the AI model that held its ground against the administration. The citizens clearly wanted less AI use in military and surveillance applications. Meanwhile, the Pentagon threatened to cancel the contract for refusing to budge on its demands. It was a US company standing up to its own government, with American citizens cheering through app installs.
Ankush Tiwari, CEO and Founder, pi-labs
AI models in military and surveillance are fraught with risks even when both the AI company and users belong to the same nation. The threat is significantly higher when nations deploy foreign AI models. The conflict between serving the national interest of the client and that of the provider’s home government can endanger the security and sovereignty of the adopting country. As every country, including India, upgrades to AI-powered military systems, the hidden risk of foreign AI models is often understated.
In December 2025, India’s national security apparatus upgraded its border management and surveillance capabilities with Artificial Intelligence (AI) and Machine Learning (ML). The upgraded tools enable predictive analysis and real-time correlation of inputs down to the district level. With this move, one of the first major technological overhauls of the MAC system in years, the Comprehensive Integrated Border Management System (CIBMS) moved toward pre-emptive intelligence rather than post-incident response.
CIBMS is a complex technological network operating behind the scenes with several overlapping layers of surveillance. It is not a single device but a dense sensor ecosystem involving ground surveillance radar, thermal imaging, fibre optic intrusion detection, and sonar. These are mounted across terrain, towers, poles, balloons, riverbanks, and underground installations, to create an invisible electronic barrier.
The recent integration of AI into such a complex system introduces risks that are difficult to fully comprehend due to system interdependencies and the unpredictable nature of AI models. Large language models and AI systems often operate as black boxes that are difficult to interpret even for their creators. When multiple tools are deployed in layered pipelines with AI integrated into the system, control over the AI infrastructure determines whether a country can safeguard its sovereign interests. In the case of foreign AI tools, there is a risk that they may not fully align with domestic priorities, thereby creating strategic vulnerabilities.
During the ongoing US–Iran conflict, Iran alleged systematic sabotage of US-made networking infrastructure, including mid-conflict hardware shutdowns that led to sudden device failures even during internet blackouts. With the advent of AI, digital warfare has evolved into AI warfare, where autonomous models can target digital systems to paralyse enemy capabilities. Countries with advanced AI companies hold an edge in this battlefield, while nations reliant on foreign technology face increasing dependency. A sovereign country cannot afford to leave its AI frontier unguarded.
Traditionally, national security focused on borders, military hardware, and energy. Today, AI models are embedded in financial systems, digital identity frameworks, cybersecurity and intelligence analysis, and critical infrastructure such as power grids, telecom networks, and logistics systems. As every digital domain becomes AI-driven, AI must now be treated as a foundational layer of national power built on three pillars: data sovereignty, model sovereignty, and infrastructure sovereignty.
Data sovereignty refers to control over data, where it is stored, who can access it, and how it is used. Countries aim to ensure that sensitive data such as citizen records, health data, and financial information remain within their jurisdiction and comply with local laws. Without this, AI systems become dependent on foreign entities that may potentially misuse or weaponise data to exert geopolitical pressure.
Model sovereignty refers to control over the AI system itself, training data, model weights, and access policies that govern fine-tuning and usage. Infrastructure sovereignty focuses on ownership and control of the entire AI stack, including compute infrastructure, models, and core software. This includes domestic data centres, locally manufactured AI chips, and independently developed AI models. Nations that control computing power and semiconductor supply chains gain a strategic advantage as the global AI arms race intensifies.
Apart from strengthening these three pillars of sovereign AI, local AI systems are also necessary to comply with privacy regulations such as the Digital Personal Data Protection Act, 2023 (DPDP Act). Another major concern is abrupt service denial. In 2025, a major American big tech company suspended services to Nayara Energy following EU sanctions linked to the Russia–Ukraine conflict. The ownership link to Russia made the company an unintended casualty of geopolitical sanctions.
As Indian companies increasingly adopt AI in daily operations, it is essential to ensure that Corporate India does not face similar disruptions. If geopolitical conflicts can lead to suspension of services or loss of data access, economic sovereignty becomes vulnerable. The need of the hour is reducing excessive dependency on foreign technology infrastructure so that India can retain independence in geopolitical strategy.
A country dependent on another nation’s AI systems cannot remain fully functional during emergencies. Sovereign AI should therefore be the next frontier in India’s national security strategy, as it will influence every other frontier of governance, economy, and defence.
-author is Ankush Tiwari, CEO and Founder, pi-labs
About The Author
