AI could leave subs with few places to hide

For over a century, the ocean has been the ultimate refuge for those who wished to disappear.

From the U-boats of the first world war to the nuclear-powered leviathans that glide through today’s deep waters, the submarine has thrived on one simple principle: stealth.

This article is the opinion of the author David Stupples, City St George’s, University of London and not necessarily that of the UK Defence Journal. If you would like to submit your own article on this topic or any other, please see our submission guidelines. 

Sound waves travel further and faster in water than light or radar waves.

This means sound is the most effective way to detect underwater objects. Modern anti-submarine warfare (ASW) is an ongoing cat-and-and-mouse game of detecting, tracking and deterring enemy submarines. With sound as the ocean’s only reliable language, ASW has primarily been a contest of listening. But the game is changing. Advances in artificial intelligence (AI), sensor networks, and autonomous vehicles are eroding the acoustic monopoly that submarines once enjoyed.

A new generation of tireless, networked and increasingly intelligent machines is beginning to patrol the seas. This promises a future where even the quietest submarine will find it harder to remain unseen. As the ocean’s soundscape becomes more crowded, navies are increasingly turning to non-acoustic methods. These technologies detect the effects of a submarine rather than its noise. Satellites equipped with synthetic aperture radar can detect subtle ripples and temperature gradients on the sea surface caused by subsurface movement.

Until recently, magnetometers, which can measure the minute disturbances a submarine creates in Earth’s magnetic field, were constrained by physics and sensitivity limits. The magnetic anomaly detectors used for ASW could only operate effectively at low altitude and at short range. Emerging quantum magnetometers, which make use of the strange science of quantum mechanics, promise orders-of-magnitude improvements in sensitivity. They could, in theory, detect the presence of a steel hull tens of kilometres away, especially when deployed in swarms aboard uncrewed aircraft or sea surface vessels.

A technique called distributed acoustic sensing (DAS) could turn ordinary undersea cables – primarily used for internet traffic – into vibration sensors. It works by measuring subtle changes in strain in the cables’ optical fibres. Through DAS, a single transoceanic cable could, in effect, become an enormous underwater microphone (hydrophone). In principle, this would allow a submarine crossing a major ocean basin to be detected by subtle pressure waves recorded in the fibres beneath it.

Autonomous vessels

At the heart of the revolution in ASW are uncrewed surface vehicles (USVs). These autonomous vessels range from small, solar-powered craft to large, long endurance ships capable of spending weeks or months at sea. Unlike crewed ships, USVs can be built cheaply and in large numbers. Armed with sonar, radar, magnetometers and communications links, they are the mobile nodes of an ocean-scale sensor network that can listen, learn and adapt in real time.

The US Navy’s Sea Hunter, an autonomous trimaran, has demonstrated its ability to track a diesel-electric submarine for extended periods without human intervention. In the UK, the Royal Navy’s Cetus project and its experimental uncrewed fleet at Portsmouth are exploring similar ideas. But it is the integration of AI with autonomy that reshapes the picture. A single USV, even a sophisticated one, can only observe a small patch of ocean. A swarm of hundreds, each communicating via satellite, laser, or acoustic link, can share information and act cooperatively.

AI is a game changer

AI does things that human operators and legacy systems cannot. It fuses data from multiple sources into a coherent picture. A single acoustic anomaly may mean little, but when combined with other data, it may form a high confidence detection.

AI operates continuously and without fatigue. Persistence is vital when hunting for the fleeting signature of a submarine designed to operate silently for weeks. And by learning how submarines navigate, avoid detection, and exploit environmental features, algorithms can forecast likely positions and movements. This could prompt ASW to move from being primarily reactive to predictive – a shift comparable to how meteorology evolved from observation to forecasting.

Through these capabilities, AI could move from simply assisting detection to orchestrating it. Humans are not leaving the loop, however. The role of human operator is shifting from hands-on detection to oversight, strategy, and what’s known as trust management. Trust is a key challenge: in this context, it’s about ensuring human decision makers understand what AI is doing and why it recommends certain actions. Navies are therefore investing heavily in explainable AI – systems that can account for their decisions – and robust communications systems that allow human operators to intervene when needed.

A connected ocean

By the 2030s, the world’s oceans may become as transparent to sensors as the skies became to radar in the 20th century. With help from AI, multiple transmitters and receivers – mounted on ships, aircraft and USVs – will be able to triangulate the positions of submarines in real time. Swarms of autonomous underwater vehicles – relatively small robotic drones – will patrol close to shore, relaying data to surface craft. Satellites will flag anomalies for local sensor networks to investigate. And the fibre-optic infrastructure that spans the seabed may double as a global array of undersea microphones.

For now, such a vision remains technically ambitious. The ocean is extraordinarily complex: temperature gradients, salinity layers, and seabed topography all distort signals and confound algorithms. But with every incremental improvement in AI modelling and computational power, those obstacles shrink. As detection grows more sophisticated, so too will the submarines. The future may see submarines using propulsion systems and materials in their hulls that leave minimal thermal or acoustic signature. Decoy drones could be used to confuse detection systems.

Some analysts predict that submarines will operate deeper and slower to evade wide-area surveillance. A shift towards autonomous undersea drones that can saturate defences through sheer numbers is also possible. The strategic implications are profound. Submarines have long been the cornerstone of nuclear deterrence and covert power projection. Their ability to vanish beneath the waves gave nations second-strike capability (the ability to retaliate after absorbing a nuclear attack) and freedom of manoeuvre.

The result of AI-driven transparency could be greater stability – reducing incentives for surprise attack – or, paradoxically, new instability as nations race to preserve secrecy.

The submarine will remain a formidable weapon, but it will no longer move unseen. The ocean, once humanity’s final hidden frontier, is becoming transparent to the eyes of machines.

David Stupples, Professor of Electrical and Electronic Engineering and Director of Electronic Warfare Research, City St George’s, University of London. This article is republished from The Conversation under a Creative Commons license. Read the original article.