Why India Needs a Sovereign Military AI Layer

On 23 October, Defence Secretary Rajesh Kumar Singh said India will build its own AI layer for the armed forces. His reason was the data. AI used for decisions, targeting and sensor fusion handles information so sensitive that, in his words, “we have to develop our own AI layer for our military.”

I think he’s right. I also think data is the smallest part of the problem.

The minutes

India has spent heavily on weapons over the last few years. ORF puts the rise in capital budgets at about 52 percent between 2019–20 and 2024–25, and its July brief argues the systems connecting those weapons haven’t kept up. The Army chief said as much last November: “This is one field in which we are not prepared.” (PTI)

What does that look like on the ground? A camera picks something up. Someone watches a screen and calls it in on the radio, someone else decides, and by the time anyone fires, a few minutes have gone by that didn’t need to.

A few minutes matters a lot. Ukraine’s Delta system brought the time from spotting a target to firing on it down from at least 20 minutes to three to five, according to a battalion chief of staff quoted by The New York Times. In May the US Army Secretary told the Senate that Delta puts every drone, sensor and shooter on one network and the US Army’s doesn’t. The Pentagon wants more than $2 billion for C2 software in 2027 to fix that.

Who’s in control

Europe is the part I keep coming back to. NATO adopted Maven, an American system, in 2025. France has since built its own, Arcadia, and the Dutch want a replacement within two years (Euronews). These are America’s closest allies, and they’re already trying to get out of a system they only just adopted.

I understand why. This layer knows where every asset is, what every sensor has seen and which options a commander looked at. A compromised radar gives away a view. A compromised decision layer gives away how you fight.

Updates are the other problem. Whoever ships the patches decides when the system changes, and that’s fine until you’re in a crisis and need it to behave a certain way.

There’s doctrine in the code as well. To rank options you have to write down priorities, rules of engagement and terrain. Train that on someone else’s wars and it’ll get ours wrong, probably when we can least afford it.

Terrain alone makes the case. Much of the war in Ukraine is fought over open, flat farmland. NATO’s eastern flank is mostly lowland plain. India has to defend high-altitude passes along the LAC, the Thar desert, the jungles of the north-east and the marshes of the Rann of Kutch. Line of sight, how far a drone can fly in thin air, how long it takes to move a unit: all of it changes from one to the next.

Where India stands

We’ve done this once. In Operation Sindoor, Akashteer, IACCS and the Navy’s Trigun gave all three services a single air picture, and it was Indian software. Singh pointed to something else we can build on: old air defence guns, fitted with radars and networked, are useful against drones again and far cheaper than a missile costing ₹5–10 crore.

Air defence is one layer, though. ORF says the gap is widest at brigade level and below, which is exactly where the minutes get lost.

So here’s what worries me. India’s forces are going to get an AI decision layer one way or another. Everyone is. What I don’t know is whose it will be when they actually need it. If we don’t build our own, the ready-made options are built, maintained and updated abroad.