By David Wojick
The Trump administration has paused offshore wind construction citing potential threats with these facilities to national security. There is little on the nature of these threats. Having done what the Navy calls vulnerability assessment of naval bases, I have taken my own whack at it.
My speculative candidate for a really big threat is an autonomous submarine armed with torpedoes or cruise missiles hiding among the giant monopiles. Call it an AI sub. There is a lot of talk about radar interference, but I am thinking sonar. Spotting this sub with sonar in a big offshore wind facility might be very difficult.
My example is the Dominion Virginia project, for two reasons. First, it lies just off the world’s biggest naval base at Norfolk. This is an obvious target. Second, I have studied this wind facility more than any other.
The Dominion facility is just under 200 square miles covered with about 180 monster monopiles supporting the huge wind towers. Each monopile is a steel tube about 30′ in diameter and extends from the sea floor to the surface. That is a lot of underwater steel spread over a large area.
My conjecture is that sonar will not work well in this forest of giant steel tubes, if at all. Surely the acoustic signal will be heavily scattered. Given that the AI sub is unmanned, it can be very small, perhaps not much larger than the weapons it carries. It could even stand on end right next to a monopile. There may also be sonar stealth materials available.
Sonar detection of AI subs might well be impossible. An alternative technology would be a swarm of much smaller AI drone subs that attacked and blew themselves up when they reached their targets. These could be even harder to detect.
Getting AI subs to the offshore wind site could be simple. There are major shipping lanes right next door. In fact, several had to be shifted to make room for the site. A properly configured freighter could drop off the subs as it slowly passed in the night. Thus, their propulsion requirements might be small.
Norfolk home ports a reported 75 warships, including a whopping four aircraft carriers. Successfully attacking this concentration would certainly be a blow to naval operations.
Note that I am not saying this threat is likely. Likelihood is the topic of threat assessments. Vulnerability assessments look for worst case plausible scenarios without regard to likelihood.
The AI Sub threat is certainly plausible. The U.S. Navy and China have both conducted drone sub research and development. For example, see here for more information.
I do not know what plausible targets lie off of the other offshore wind sites, but no doubt there are some. This includes civilian sites as well as military. Civilian targets might range from interstate bridges to entire cities.
It is up to the Trump administration to carefully assess the threat of tiny AI subs hiding among the massive steel monopiles of offshore wind facilities. Results might range from banning facilities in certain areas to developing detection and defense systems where development is allowed, if that happens. In some cases, the monopiles are already there. Perhaps they should be removed.
The scale of the AI sub threat is clearly far greater than the threat of radar clutter. It needs to be taken very seriously.
That’s why nations with any kind of marine presence have navies, and air forces…and the will to deploy them…
The issue with submarine drones is power longevity.
A friend once asked the question, “why aren’t drugs smuggled by sea drone?”
One obvious option is an underwater stealth drone. But underwater drones have to surface regularly to recharge, if they are powered by batteries. Solar energy is only available around midday, so you would have to surface for at least a few hours every day. Even if the drone went into near shutdown, it would still have to surface sometimes to recharge.
A surface drone similar to the small drones Ukraine are using to attack Russian ships in the Black Sea might work, but it wouldn’t be that stealthy it could linger for weeks or months.
A small nuclear power plant would circumvent this issue. But even with nuclear drones you encounter some serious issues. Nuclear thermal plants like the Pioneer Space Probe or as used in some remote Soviet lighthouses would work, but the isotopes for nuclear thermal, which have to be synthesised in a reactor, are hideously expensive to create, so mass production would be a challenge. Miniature nuclear fission reactors would be cheaper, but a miniature drone wouldn’t have enough shielding to prevent neutron bombardment of surrounding seawater. Sooner or later someone would notice the sea water near a sensitive site was radioactive.
I think the AI attack on the USA will be far more subtle. People, especially teenagers, already spend hours with TikTok and other AI engines operated by hostile powers. What if one of these engines was programmed to offer helpful life advice, like how to get a job, how to get promoted? Then you would have a drone, someone utterly dependent on their AI “friend”, who was helped by AI to rise way beyond their level of competence, who could be manipulated into causing harm. Imagine trying to fire up a wartime economy, only to have thousands of these AI drones in management positions simultaneously manipulated into throwing metaphorical sand into the gears of your economy? I’m not saying they would be intentionally disloyal, but if they depend on their AI crutch for every aspect of their lives, it wouldn’t be hard to persuade people who don’t really understand what they are doing to launch a sophisticated multi-agent takedown of productive capacity or wartime resupply.
The Dominion site is just 30 miles from the Naval Base so little travel is required. If they launch cruise missles no travel. No need to recharge.
They still need to listen to their HQ, for years. If they aren’t going to wait, what’s the point?
Yes but it takes little energy to listen.
An uncrewed submarine would have nowhere near the power requirements of a typical naval warship, and could easily be designed to charge while submerged via snorkel. The design practices evolved by the Argo buoy system would go a long way toward producing a world-wide network of stealth submarines just waiting for the balloon to go up, so to speak.
Could all those pylons provide an opportunity to install sensor arrays? Modern submarine detection methods go beyond just sonar, such as blue-green laser and magnetic anomaly detection.
Yes that is a possible defense action.
Thinking of this some more, there indeed could be advantages for detection. Even with sonar, a normal pattern of sound waves–a field–would be set up by the pylons with nothing between them. As soon as another object is inserted into this field the disruption of the standard pattern could be recognized, as in ‘I detect a disturbance in The Force, Luke’. Computational power today is such that such a disturbance could be determined quickly enough to be effectively real-time.
If one considers dense electrode arrays of a resistivity system like E-Scan for example, the position of an intruder could be determined in 3D as well, with sufficient accuracy for targeting. I can see where these things could be made into effective sensor arrays that make them the opposite of a problem–which means it would be a good idea to keep discussing what a problem they are while the algorithms are being quietly developed and defensive targeting hardware installed.
(I enjoyed the sarcasm of Spy vs Spy in the early 60’s…)
fish?
I think you’ll find that a suitably appointed judge would prohibit this usage.
Japanese WW2 Type A 24 metres long 2 torpedoes 2 man crew attacked Pearl Harbour 1941 and Sydney Harbour 1942. Developed in Koryu with a range of 450 miles 2 torpedoes and 2 man crew at 18 metres long. Then they had the Kaiten manned suicide torpedo. None of these were particularly successful.
The Italian Navy developed the Chariot a 2 man crewed torpedo which they used to sink the Battleships Queen Elizabeth and Valiant in Alexandria Harbour.
The Royal Navy had the X Class which were used against Tirpitz and in the D Day landings. Also Singapore Harbour was attack.
The crews of the British and Italian submarines were supposed to survive a mission not all did.
I don’t think the US Navy used them in WW2. But in the 80 years since small submersibles have come a long way and are in use every day commercially
Being uncrewed makes AI Subs much smaller.
Could AI undersea drones would be able to tap onto power generated by the wind tower by induction?
They could bury themselves in the mud next to the cabling or even on individual tower poles to become stealth.
Eventually, even if the majority were detected, the ‘sleepers’ could arise and do whatever they were designed to do.
Good point re digging in.
Don’t long distance connections use HVDC connections which means no induction charging?
I think only one site uses HVDC.
If they can dig in, they can use tidal forces to recharge batteries.
So they’ll all break down within 6 months. Self extinguishing weapons. Maybe they should stick to induction.
High voltage insulated cables require grounded shielding. That should contain the induction, as I understand it.
Mad Ed in the UK keeps adding windmills to the English Channel.
I guess it would do no good to send Mad Ed a copy of this article.
They could all have a timer, set for months and years, going off at the same time, not needing AI. The actual propulsion could be an encased rocket motor, making either a quick or slow run towards a target, passing by or fixed. The battery could be small for powering the AI and sensors. A kill timer could flood and disintegrate everything, and nobody would know it existed. The battery could be an external block of aluminum, which uses the salt water as electrolyte. Also instead of surfacing for instructions it could use a thin floating wire adjacent to the pylon, which is easily disguised as seaweed. It is not all that difficult for a dedicated development team.
I’m thinking that Russia’s new nuclear propulsion Poseidon technology sort of makes all this a mute point.
How so?
From my understanding, the Poseidon system is a nuclear torpedo that also utilizes the nuclear propulsion system that Russia just demonstrated with a cruse missile that has an unlimited range. Same propulsion system that Russia uses on their nuclear powered subs, except much much smaller. Apparently they can be stationed off the coast line for an indefinite period of time and launched to target remotely. Might all be a lot of bull s**t, but then Russia has been showing some stuff lately that is only a dream for the USMIC.
The drones go slow, and then fast, for a long way.
The US Navy reports on it here:
Russia Introduces a New Category of Naval Strike Weapon | Proceedings – January 2026 Vol. 152/1/1,475
This is a source with relevant expertise and a known interest in the subject.
Of course, they have a bias towards exaggerating potential threats to ensure they get continued funding. But, unlike most research institutes, the US Navy has little fear of defunding, at the moment.
Surely the network of inter array cables could make a very good sub detection system without the need for any active sonar?
There are reasons for and against offshore wind, largely against due to the cost and intermittancy, but hiding subs should be the least of them.
How would the cable system detect anything?
Why do you consider this unimportant? Is losing a war less important than intermittency?
The fibres within them can work as passive acoustic sensors.
In April 2025 Wind Europe warned that physical and cybersecurity attacks on “Europe’s critical energy infrastructure are a reality” and that the threat level had increased since Russia invaded Ukraine. The Baltic Sea had at least 6 suspected sabotage incidents aimed at electricity and internet cables since 2022 and11 known undersea cables had “been taken out since 2023”
data apply cyber
They also pointed out that “Wind turbines are packed with electronics and sensors which gather a wealth of operational data….This makes wind turbines potential targets for cyber attacks”
“To ensure energy system reliance it is critical governments apply cybersecurity and data protection as pre-qualification in wind energy auctions and public procurement strategies”
Wind Europe 25/4/25 ‘New collaborations to ensure physical and cyber security of wind turbines’
I thought about this years ago. Every time a blade passes in front of the mast, it emits a low frequency pulse. Bad for whales and other creatures, and obviously bad for detecting subs. I worked as a consultant for a company that developed a means of measuring this noise from wind turbines. They gave people living nearby land based turbines a handheld to activate when they felt adverse effects, such as getting a headache, nausea, etc. Even though humans can’t actually ‘hear’ at this frequency, there is negative health effects they were able to correlate and measure.
There is an obvious problem with the human push button devices. If the human was intent on sending a highly correlated signal, all they would have to do is listen for the wind blowing around their house. If it is windy, then the turbines will be spinning and hence it’s time to press the button. This could also be triggered if they could see the turbines spinning.
And of course, you can’t do a test where the human is placed in a fully isolated box so that they can’t hear the wind nor see the turbines, if you did, then you are also likely to be blocking the very sound that they are objecting to.
If infra-sound was incapable of causing harm, then why have the military investigated it’s use as a directed energy weapon? I think it’s safe to say that the energy levels are not harmful in the sense of material damage but they are not without harm. I would not like to live anywhere near one, hearing/feeling a pressure pulse from the blades is not on my list of things that I want to permanently study from my home.
An interesting thought but the primary reason offshore wind farms shouldn’t be built is because they don’t work.
And they are ugly.
Agree with the conclusion that underwater “AI drones should be taken seriously.”
But for many reasons not cited here. Everything that floats or is lying in or anchored in the ocean is vulnerable. I see many options for defenses in your thread, but every reason to believe that the best the U.S. can do is develop the most sophisticated microprocessors. first and maintain that advantage in perpetuity. What is vulnerable? (AI Query):
How U.S. Navy and Air Force defend our 11 aircraft carriers (four of which are in U.S. ports) is a also a big question.
“The Dominion facility is just under 200 square miles covered with about 180 monster monopiles supporting the huge wind towers. Each monopile is a steel tube about 30′ in diameter and extends from the sea floor to the surface. That is a lot of underwater steel spread over a large area.”
Actually it is a tiny amount of steel spread over a large area. The ratio of steel to water given the numbers above is about 1:10 million. Futhermore the range of an active sonar system is only a couple of kilometres meaning that most of the facility would be beyond the range of such a system and thus not causing interference. Passive sonar has a range of about 100km but it relies on detecting moving objects and so wouldn’t be able to detect the stationary AI subs that David is worried about irrespective of whether or not the dominion facility is built.
Lots of talk about how to make underwater AI drones. Bigger question is “What problem do they solve?” I can’t think of a situation that doesn’t have a cheaper, easier or more effective solution.
If the towers of obscenities can confuse sonar, what does that do to whales?