Kent’s AI Border Towers: £21m Spent as 153,000 People Arrived by Small Boat — Now the Home Office Plans to Spend £120m More

Walk west along the shingle at Hythe and, above the beach, an unusual structure looks out over the Channel. Roughly 18ft high, it carries radar and electro-optical and thermal cameras (pictured). This isn’t bog-standard CCTV. It’s an Anduril Maritime Sentry, built to use artificial intelligence and “sensor fusion” — combining information from different sensors — to detect, identify, classify and track objects at sea. Anduril says Maritime Sentry can detect surface vessels at more than 15km (or more than 9 miles) and that one operator can manage more than ten towers.

Hythe is one piece of a coastal network. Research by Samuel Storey and the Migrants’ Rights Network identified eight towers understood to be operating between Hastings and Margate in April 2025: Fairlight Coastguard Station, Dungeness Lighthouse, Hythe Ranges East Lookout, Shorncliffe Army Camp, Capel-le-Ferne Cliffs, Dover Maritime Rescue Coordination Centre, Walmer and Kingsdown Golf Club and Ramsgate Port. Two further installations, at Jury’s Gap and Hythe Ranges West Lookout, had apparently been dismantled or relocated. The Home Office doesn’t publish a current tower-by-tower operational inventory, so we can’t say with certainty how many remain active today.

But the tower is only the bit we can see.

Behind it was a much larger Home Office service called the Common Operating Picture and Command Interface, or COPCI. The latest published contract record shows Anduril Industries UK Ltd was awarded £21,086,291.12, with the contract running from 22 June 2022 until 22 June 2026. It was a managed service rather than the Government simply buying individual towers.

And it wasn’t openly advertised.

The official procurement record says the contract was awarded through a “negotiated procedure without prior publication”. That procedure can legally be used in certain circumstances — for example, where only one supplier can provide the requirement or extreme urgency applies — so the procurement route isn’t, by itself, evidence anything improper happened. What it does mean is that there wasn’t an ordinary advertised competition before Anduril got the contract.

How did the towers get onto Kent’s coast?

Some of the most interesting questions start on dry land.

The installations at Hythe Ranges and Shorncliffe sit on Ministry of Defence property. The disclosed land records behind this investigation show the sites were made available through Military Aid to the Civil Authorities, better known as MACA. That mechanism allows Defence to support bodies including Border Force and can include making the Defence estate or facilities available for operational purposes.

Yet the papers concerning Hythe and Shorncliffe say the MoD hadn’t received ground rent, licence fees, lease payments, wayleaves or easement payments from Anduril during the four-year occupation, and that the department only later moved to regularise matters through a licence. A £14,000 charge concerning an electrical connection at Shorncliffe was identified separately. That doesn’t automatically establish wrongdoing: MACA may have provided the authority for occupation while a separate property agreement was being pursued. But after four years, taxpayers are entitled to ask why the property arrangements took so long to regularise.

That question becomes more interesting when set against the surveillance procurement itself. The succeeding Home Office specification expressly places responsibility on the supplier for owning, siting, installing, securing and operating the hardware, including making the commercial arrangements necessary to use land. In other words, land isn’t some minor administrative afterthought in this type of service. It’s part of delivering it.

Then there’s the bigger question.

Did it work?

What happened while the £21m contract was running?

The timing matters because Anduril didn’t arrive at the beginning of the small-boats phenomenon. The route was already expanding at extraordinary speed. Home Office figures show 299 people arriving in 2018, 1,843 in 2019, 8,466 in 2020 and 28,526 in 2021. By the end of June 2022 — before Anduril had completed even its first full month — another 12,747 people had arrived.

The published contract began on 22 June 2022. That day alone, Home Office operational data recorded 98 people arriving on three boats. Between 22 and 30 June there were more than 1,000 arrivals. August then became the biggest month recorded at the time, with 8,631 people, and the final 2022 total reached 45,774.

The following year was very different. Arrivals fell to 29,437 in 2023, before rising to 36,816 in 2024 and 41,472 in 2025. So the four-year contract period doesn’t produce a tidy political narrative in either direction: numbers plunged, then climbed for two successive years, before falling substantially during the first half of 2026.

We’ve gone one stage further and aligned the Government’s small-boats time series as closely as the published data allows with Anduril’s exact 22 June 2022 to 22 June 2026 contract window. Using final Home Office annual figures where available and provisional operational data for the partial years at either end, the result is approximately 153,000 people arriving in about 2,900 boats during the lifetime of the published £21.086m contract. We’re deliberately rounding those figures because the Home Office warns that daily operational data can later be revised and that its quarterly statistics are the final authoritative series.

It would be tempting at this point to write: £21m spent, 153,000 people arrived, therefore the system failed.

But that isn’t what the evidence proves.

An arrival can mean the surveillance worked

The Home Office definition of a small-boat “arrival” includes not only somebody who reaches a British beach. It also includes someone detected in the Channel by UK authorities and subsequently brought to the UK. That distinction turns the obvious interpretation of those 153,000 people on its head.

Suppose an Anduril sensor – such as this one pictured at Hythe Ranges – spotted a dinghy miles offshore, the common operating picture alerted Border Force, a rescue vessel intercepted it and everybody was brought safely ashore. Every person on that boat could still enter the Government’s arrival statistics. An arrival, therefore, can be evidence of a successful detection rather than a missed one.

The Government’s own latest assessment makes that especially clear. The Border Security Commander says 99.9% of small boats reaching UK waters since 2022 have been interdicted, with those rescued then secured and brought ashore. It describes Intelligence, Surveillance and Reconnaissance — ISR — as a critical extra layer for preventing beach landings and helping identify and prosecute offenders. It also says physically trying to stop heavily overloaded boats once they’re at sea can create an immediate risk of mass casualties.

So 153,000 arrivals cannot honestly be called 153,000 Anduril failures.

That doesn’t let a £21m contract off the hook. It tells us which performance question we ought to be asking.

The smugglers changed the arithmetic

Counting people alone also misses one of the most important changes during the contract.

In 2022 there were 1,109 recorded boats, carrying an average 41 people each. In 2023 the average rose to about 49. By 2024, 36,816 people arrived on 695 boats, giving an average of 53. Then in 2025 the number of boats actually fell to 672, yet arrivals increased to 41,472 because the average load climbed again to 62 people per boat.

That’s adaptation in action. The National Crime Agency says organised crime groups increasingly used tactics such as “taxi boats” and larger passenger loads. The Border Security Commander similarly reports simultaneous launches designed to stretch French enforcement and says average loads reached 63 across the 2025/26 financial year. Fewer successful boats, in other words, needn’t mean proportionately fewer people.

The first half of 2026 shows why simple before-and-after claims remain dangerous. By June, arrivals were 41% lower year-on-year, according to the Border Security Commander, but the Government itself remains cautious about attributing that fall to any single intervention. Weather, enforcement in France, equipment seizures, intelligence, returns, prosecutions, migrant demand and changes in smuggling tactics are all moving at once.

So what did £21m actually buy?

Here’s where the value-for-money investigation hits the wall.

In a March 2026 Freedom of Information decision concerning an Anduril tower at Jury’s Gap – near Lydd, the Information Commissioner recorded the Home Office’s explanation that there were no individual tower costs because the department had purchased a managed service rather than separate towers. More importantly, the Home Office told the Commissioner that there were monthly reports and meetings with Anduril concerning the effectiveness of the whole system.

That means performance was being monitored.

What the public still needs is enough aggregated information to understand it. How many small boats did the Anduril system detect? How often was it the first system to make a detection? What proportion of alerts were genuine? What was availability or uptime? How often did a detection result in operational tasking? Did detection or response times improve? How many uncontrolled landings involved craft that hadn’t been detected? Were contractual performance targets met, and were service credits or penalties ever applied? None of that requires publishing a map of surveillance blind spots or giving smugglers sensitive technical specifications.

Without those figures, the public can see the price but can’t properly measure the product.

Can we put an average price on the arrivals?

We can calculate something useful — but it needs the right label.

Divide Anduril’s £21,086,291.12 published contract value by approximately 153,000 recorded arrivals during its contract period, and the result is roughly £138 of contract value per person recorded arriving. Divide it by approximately 2,900 recorded boats, and it works out at about £7,250 per boat. Across the same period, the crude average was roughly 53 people per boat.

Those are not £138 per failed person or £7,250 per failed interception. Calling them that would be wrong, because the Home Office counts intercepted people as arrivals and says 99.9% of boats reaching UK waters since 2022 have been interdicted. The defensible description is simply Anduril contract value per recorded small-boat arrival and per recorded boat during the contract period. Whether those ratios represent good or bad value depends on the performance information the Home Office says exists but which isn’t presently available to make that assessment.

And we can’t sensibly repeat that calculation by adding every maritime contract together. Some services overlap, some perform different functions, some are merely planned rather than awarded, and one proposed contract was swallowed into another. Adding the headline values and calling the result “the cost of failing to stop the boats” would double-count some requirements and pretend proposed expenditure had already been spent.

Then the money gets much bigger

Anduril isn’t the only surveillance expenditure in the Channel picture.

The Home Office awarded PAL Aerospace Ltd £33,926,306 for a strategic fixed-wing ISR aircraft managed service running from 29 October 2024 to 31 January 2026. Like Anduril’s contract, that award used a negotiated procedure without prior publication. A separate £4m Digital Maritime Security Capabilities contract was awarded to Carbon60 to integrate maritime data sources into a coherent maritime-domain-awareness picture, running from July 2025 to June 2027. The aircraft flew out of Lydd.

There’s also the next generation of airborne surveillance. The latest procurement trail puts the Tactical Remotely Piloted Aircraft System — effectively a managed drone service — at an estimated £37.5m. The requirement is for aircraft capable of detecting, identifying and tracking vessels and individuals, transmitting information to command-and-control systems and providing footage capable of supporting criminal and financial investigations. Earlier market-engagement papers envisaged 5,000 on-task sensor hours each year.

Then comes the big one.

The Home Office’s latest Coastal Maritime Intelligence, Surveillance and Reconnaissance Service has an estimated maximum value of £99,833,333.33 excluding VAT — £119.8m including VAT. The proposed contract would initially run from 1 August 2027 to 31 July 2030, with options potentially extending it to July 2032.

The scale needs explaining carefully. An earlier Home Office proposal envisaged spending up to £12m excluding VAT on a Land-Based ISR and Common Operating Picture service covering the Dover Strait from Ramsgate to Hastings. Its documents describe the exact problem: inflatable vessels typically around eight metres long, often carrying 40 to 60 people, usually without AIS identification equipment and presenting a very small target to conventional radar. The supplier would provide AI-assisted sensor fusion and 24/7 surveillance.

But that £12m mustn’t be added on top of the £99.833m. The Home Office subsequently merged the Land-Based ISR requirement with the Joint Maritime Security Centre’s Mobile Maritime Domain Awareness requirement to create the single Coastal Maritime ISR procurement. The successor is intended to combine radar, AIS and electro-optical/infra-red sensors, autonomously detect and track small boats and other vessels, and feed near-real-time tracks into Royal Navy and Home Office systems.

The same caution applies to the financial picture as a whole. PAL aircraft, Anduril towers/software, digital maritime-data integration, drones and Coastal Maritime ISR aren’t five versions of the same invoice. Their periods and functions overlap, while the £37.5m drone service and £99.833m Coastal service are estimated future procurements rather than money already paid. A big combined number would look impressive on a headline, but it wouldn’t be forensic accounting.

And then the procurement trail breaks

This is the point where the chronology becomes awkward.

Anduril’s published COPCI contract ended on 22 June 2026. The earlier £12m Land-Based ISR replacement had been provisionally timed to start on 23 June 2026 — literally the following day. But that procurement was subsequently consolidated into Coastal Maritime ISR, whose current planned start isn’t until 1 August 2027.

That leaves more than a year between the published end of Anduril’s contract and the expected commencement of the expanded replacement.

We’ve searched the public procurement trail extensively for an Anduril extension, bridge contract or replacement award covering that gap and haven’t located one. That doesn’t establish that no arrangement exists. Security contracting can involve information being withheld, and previous Information Commissioner proceedings involving Home Office surveillance contracts have confirmed that national-security exemptions may apply. But if the existing capability remains operational, the entirely reasonable question is: under what commercial arrangement is it currently being supplied and paid for?

There’s another loose end. The June Coastal Maritime ISR notice said a formal UK4 Tender Notice was expected on 3 August 2026, with expressions of interest due by 15 September. As of 22 August, the public record we’ve been able to locate still shows the procurement as planned; we haven’t found a separately indexed UK4 notice. That may have an entirely mundane explanation — suppliers are being routed through a Government dynamic market, the Home Office’s Jaggaer procurement portal and non-disclosure agreements — but the Home Office should be able to confirm whether the tender has been issued and under which notice reference.

Set the scoreboard before another £119.8m

This is where the Government has an opportunity to do something it apparently didn’t do publicly with the Anduril contract: establish the value-for-money test before the next service begins.

We already have the baseline. We know the number of people arriving, the number of boats, average occupancy, uncontrolled landings and — from April 2024 — French prevention figures. The Border Security Commander says more than 20,000 crossing attempts were prevented by France in 2025/26, and more than 40,000 since the Command was established. It also acknowledges that assessing impact is difficult and that further evidence is needed before long-term changes in organised crime groups’ behaviour can be demonstrated.

The missing half is the surveillance performance itself.

Before potentially committing up to £119.8m including VAT, the Home Office could publish security-safe measures covering detection rates, uptime, first detections, false alarms, response times, operational tasking, uncontrolled landings and achievement against contractual KPIs. Then, when the contract has run for a year, Parliament and the public could compare the result with the baseline instead of trying to reverse-engineer performance from arrival figures that were never designed to measure surveillance effectiveness.

The £138 per recorded person and roughly £7,250 per recorded boat figures tell us what Anduril’s £21.1m price looks like alongside the traffic recorded during its contract. They don’t tell us whether the service succeeded or failed. For that, we need the monthly performance information that the Home Office has already confirmed existed.

And that’s where this investigation ends — for now.

If the Government can’t show taxpayers, in aggregated and security-safe terms, how the £21m system performed, how will anyone know whether a successor worth up to £119.8m represents value for money?

The Shepway Vox Team

Dissent is NOT a Crime