Undersea innovation at the speed of the threat

On the 9th of April, the UK Government announced it had exposed and disrupted a covert Russian submarine operation in and around UK waters. 

It had lasted more than a month, with over 450 flight hours clocked by submarine hunting aircraft, several thousand nautical miles covered by HMS St Albans, and 500 British personnel involved. And yet it unfolded in complete silence.

This was the latest in a series of undersea provocations from Russian adversaries. Over the past two years, the UK has detected a 30% increase in Russian vessels posing a threat to its waters.  

That the UK thwarted this latest operation is a testament to the armed forces involved. But the operation illuminates uncomfortable vulnerabilities for an island nation. The cables and pipelines targeted carry half the gas that heats British homes, 99% of international telecoms and data traffic, and trillions of pounds of global trade every day. They are not abstract strategic assets. They are the infrastructure of everyday life – and they sit, largely unguarded, on the seabed.

The gap isn't technical. It's about tempo.

Traditional defence acquisition models were designed for a world in which threat evolution was measured in years, sometimes decades. That world is gone. At Thales, the question we tackled at Undersea Defence Technology (UDT) is not just "what technology do we need to enable navies to detect first, decide faster, and control the underwater battlespace?" – it is "how do we develop, assure and deploy it at the speed the threat demands?" There are three areas where the answer must change.

Go to sea to understand what you're building for

The ocean is not a laboratory, and it punishes assumptions made only in one. Glint, haze, variable sea states, rapidly changing atmospheric conditions – these are not edge cases, but standard operating conditions. Maritime performance cannot be reliably predicted from controlled environments, meaning systems developed and validated only in those environments arrive in the field with an unknown gap between demonstrated capability and real-world performance.

This has direct implications for how novel sensor technologies are evaluated. Polarimetric cameras, SPAD imagers, event-based sensors – each promises compelling advantages on paper, but only real trials in real conditions reveal how those advantages translate and where they break down. Thales’ coastal field testing revealed that no single sensing modality is sufficient across all maritime environments – instead the greatest capability comes from fusing complementary strengths.

The same logic applies to AI. The potential of Machine Learning (ML) in the underwater battlespace is well understood, but ML outputs are only as reliable as the inputs they are trained against, and models built on idealised data fail in the field in ways that are difficult to anticipate. Robustness requirements must be derived from actual operational conditions – including the environmental variability and platform dynamics that affect what the system sees before the algorithm can get to work.

Going to sea early, even with imperfect systems, is not a shortcut. It is the discipline.

Be ambitious about what uncrewed systems can do

The coverage problem in contested waters cannot be solved by crewed platforms alone. The ocean is too vast, the missions too persistent, and the risks too high. But the answer is not simply to add UxVs. Instead, the question is how ambitious we are prepared to be about what those platforms can contribute – and whether we are designing them from the outset as genuine force multipliers rather than narrow-mission assets.

Consider what becomes possible when the sensor architecture is designed for breadth rather than a single use case. By miniaturising and adapting best-of-breed sonar technology for XLUUV platforms – an area where Thales has invested heavily in self-funded research and development – a single uncrewed vessel can complete a whole series of tasks.  Without reconfiguration between missions, it can conduct passive ASW, active search operations, seabed mapping via low-frequency synthetic aperture sonar, acoustic communications, and act as an offboard sensor asset in support of crewed submarines operating nearby. That is not an incremental improvement, but a step-change in what a platform of that size and cost can contribute.

The force multiplication argument matters here too. The real value of uncrewed systems is not what they do in isolation. It is what they unlock for the crewed platforms operating alongside them – feeding a richer, more persistent common picture that enables faster, more confident decisions at the command level.

Build trust into the iteration cycle, not as a gate at the end

Speed and trust can feel like opposites in defence development. The instinct is to assure everything before deploying anything, yet the modern adversary is not pausing while the assurance case is completed.

The "move fast and break things" approach that works in commercial software development is not transferable to the North Atlantic. Broken capability in a contested environment creates a coverage gap that a sophisticated opponent will identify and exploit. Yet waiting until every test is complete is equally untenable when the threat is present and accelerating.

The answer, instead, is a different philosophy of development entirely: one that builds rigour into the iteration cycle rather than appending it at the end. This means deriving requirements from real operational conditions upfront; deploying against primary use cases with clearly understood constraints; expanding the operational envelope progressively as evidence accumulates; and using digital twins and synthetic wargaming to compress the learning cycle before hardware ever goes to sea.

This is not a compromise between speed and safety. It is a recognition that sequential gate-based development is itself a risk in the current threat environment.

The advantage is built now, or not at all

These three imperatives are not separate lines of effort, they are mutually reinforcing. Better data from real operational environments produces more trustworthy AI. More trustworthy AI enables more ambitious autonomy. More ambitious autonomy, designed for integration from the outset, delivers the force multiplication that navies actually need. And all of it depends on a development philosophy that moves at the pace the situation demands.

The underwater advantage does not accumulate passively. It has to be built, tested and deployed – and then sustained and adapted through the full lifecycle as the threat continues to evolve. The events of recent weeks are a reminder that that lifecycle is already underway, whether the capability to match it is in place or not.

The technology is maturing, the threat is unambiguous, and the strategic will is clearly present. What is needed now is the courage to close the gap between intent and delivery – to ask the hardest questions and to work through the answers together with the urgency they deserve.