British divers from the Diving and Threat Exploitation Group ran a VideoRay Defender-Viper, the small remotely operated vehicle used to find mines. Hydrographic specialists based at Devonport put a Teledyne Gavia autonomous vehicle in the water alongside American and Australian crews.<\/p>\n Lieutenant Commander Mark Wilton of the Royal Navy’s Hydrographic eXploitation Group said his team was “demonstrating true integration through the seamless use of Royal Australian Navy equipment”<\/a> when it was needed. That is a small sentence doing heavy lifting. Borrowing another navy’s hardware mid-exercise is the part that usually breaks.<\/p>\n Anduril and Ultra Maritime were not the only vendors there. Bedrock Ocean, Integer Technologies and Boeing’s Liquid Robotics ran a separate autonomy demonstration<\/a> at the same event, working concurrently out of Monterey, California and Little Creek, Virginia, with a Wave Glider on the surface acting as the relay between submerged vehicles and shore.<\/p>\n Ultra Maritime’s release carries a Keyport, Washington dateline. Keyport is a small town on the Kitsap Peninsula that has been the Navy’s undersea test site for roughly a century and still hosts one of the two Naval Undersea Warfare Center divisions. Neither company said exactly where its own portion ran.<\/p>\n The division of labor here matters, and a lot of the early write-ups blurred it. Sea Spear is the sensor. It is a deployable sonar array from Ultra Maritime, and it is the thing that actually heard the intruders and worked out what they were.<\/p>\n Seabed Sentry is the node the array lives on and the network it talks through. Per the releases carried by Naval News<\/a>, Seabed Sentry took the tracks from Sea Spear and published them into Lattice, Anduril’s battle management software, which then pushed the resulting picture into the Navy’s own counter-uncrewed-vehicle system. An operator in the Navy’s Unmanned Operations Center saw it there.<\/p>\n Anduril singled out two results as the ones it cared about. The first is that the alert went through without a person in the middle. The second is that its software chain plugged into the Navy’s existing command setup with no custom engineering on either side, which is a claim worth remembering the next time a defense integration program slips three years.<\/p>\n Lattice is not new. It has spent years running counter-drone and perimeter work in the air, fusing radar and electronic warfare feeds into one picture. Lanternfish tested whether the same stack survives an environment where the sensor is acoustic and the bandwidth is dreadful.<\/p>\n The physical object is unremarkable to look at. Each node is a cylinder 21 inches across and roughly eight feet long, with the computing and communications gear in the head and a swappable payload bay in the tail. It carries its own anchor.<\/p>\n Anduril’s published figures put the depth rating above 500 meters, payload volume above half a cubic meter, and endurance at months to years depending on configuration. Those last numbers are manufacturer claims, not measurements anyone outside the company has published.<\/p>\n The reason the cable goes away is bandwidth. You cannot push raw sonar audio through seawater the way you push video through fiber, so the processing happens on the node itself and only the finished product goes out: a detection, a track, a classification. Those compressed messages hop between nodes on acoustic modems and out to whatever is listening, which might be another node, a ship, or the uncrewed vehicle that dropped the thing off in the first place.<\/p>\n Recovery is the part that sounds made up and is not. When a node is finished, the anchor releases itself, the node floats up, and somebody scrapes the barnacles off, recharges it, swaps the payload if the mission changed, and drops it back in.<\/p>\n Sea Spear extends out of the node on a collapsing truss that Ultra Maritime says was inspired by lightweight extendable arms built for spacecraft, reaching tens of meters when deployed. It listens passively most of the time, which is the whole idea. An array that pings gives away its own position.<\/p>\n Dr. Shane Arnott, Anduril’s senior vice president for engineering and its maritime lead, described the seeding concept at the April 2025 unveiling, telling The War Zone<\/a> that “we can load up the Dive XL with about a dozen of these things” and place them covertly. Neither company has said how the nodes got into the water at Lanternfish, so that line is a 2025 sales pitch, not a description of what happened this month.<\/p>\n Here is where the reporting has to slow down. Both accounts of Lanternfish say the same thing about what was detected: medium and large-diameter uncrewed vehicles. That is the easier end of the problem.<\/p>\n Anduril’s own argument for harbor defense, in the same release, is that the threat to ports no longer looks like a submarine. It looks like a swarm of small, cheap autonomous machines, and a system built to watch for big things misses them entirely. Small is the hard case, and small is not what the companies reported catching.<\/p>\n The 500-meter figure is a rating, not a description of the exercise. A port protection scenario happens in water nowhere near that deep, and shallow water is acoustically nastier than open ocean, with reflections off the surface and the bottom cluttering everything.<\/p>\n And both narratives are corporate. Naval News and Seapower ran the Ultra Maritime and Anduril statements as press releases, which is normal practice and labeled as such, but the Navy has not published its own assessment. There is no independent scoring here, just a vendor saying it worked and an ally confirming the exercise happened.<\/p>\n When Seabed Sentry was unveiled in April 2025, Anduril said it could not name a single customer. Fifteen months later, that has not visibly changed for this specific product.<\/p>\n What has changed is the surrounding program. On May 30, AUKUS defense ministers meeting in Singapore approved the first signature project under Pillar 2<\/a>: joint development of payloads, sensors now and weapons later, that can be carried by any of the three nations’ uncrewed underwater fleets. The UK Ministry of Defence said the first capabilities are expected in service by 2027. We covered how that payload effort ties the three fleets together<\/a> when it was announced.<\/p>\n Britain is running its own competition in parallel. Atlantic Net, the first phase of the Royal Navy’s Atlantic Bastion push to watch the North Atlantic with uncrewed systems, drew 26 compliant industry proposals. Janes reported in April<\/a> that five consortia remained in contention, led by Anduril, Babcock, Helsing, Ocean Affinity and TKMS. No winner has been announced publicly.<\/p>\n That list is worth reading twice, because Helsing is on it with a rival approach. Its SG-1 Fathom is a 132-pound glider that parks on the seabed and listens with an AI trained on decades of ocean sound<\/a>, and the Royal Navy has been building programs around it. A Danish startup called Triton is scattering cheap passive microphones across the same seabed<\/a> with the same basic logic. Anduril is not the only company that noticed the cable was the weak link.<\/p>\n It does have more money than most of them. Anduril was reportedly in talks in late July to raise at a valuation near $100 billion, and it has been expanding its maritime engineering in the Seattle area, on the other side of Puget Sound from Keyport. Its Dive-XL was picked by the Navy and the Defense Innovation Unit in March, and Australia is already building the related Ghost Shark in Sydney<\/a>.<\/p>\n The engineering question Lanternfish answered was whether a wireless node could hand a track to a Navy screen fast enough to be useful. It could, repeatedly, across a six-week exercise.<\/p>\n What it did not answer is what one of these sounds like in month four on the bottom, with growth on the array and nobody overhead to check. Anduril says they come home, get cleaned and go back out. No customer has said publicly how many it wants or where they would go. Until one does, America’s wireless alarm on the seafloor is a very good demonstration with a Keyport dateline.<\/p>\n","protected":false},"excerpt":{"rendered":" Underwater listening arrays have worked more or less the same way since the Cold War. A hydrophone sits on the … <\/p>\n![\"teledyne](\"https:\/\/www.autonocion.com\/us\/wp-content\/uploads\/2026\/07\/teledyne-gavia-drone-royal-navy-1024x576.jpg\")
The sonar does the listening, the can does the networking<\/h2>\n
Cutting the cable was the entire point<\/h2>\n
The drones it caught are not the drones it is sold against<\/h2>\n
Nobody has announced a contract for it yet<\/h2>\n