AI (Artificial Intelligence) in the military

A senior systems engineer with the (UK) MoD who has decades of experience across the defense industry is highly critical about Palantir's access across data sets:

MoD Systems engineer said:
Ministers clearly have a lack of understanding of Palantir’s technology. The statements with respect to sovereign data appear to be missing the point entirely. [They’re] missing the realities of data scraping, of aggregation, and the fact that Palantir is building its own rich picture of our nation that they can use for their own ends.

Allowing a single entity, foreign or domestic, to have such far-reaching, pervasive access is inherently dangerous. How our national cybersecurity centre has allowed this beggars belief.

Whether or not the UK technically owns the data is almost irrelevant. That’s like reading a secret love letter and saying the secrets in it are safe, just because you’ve promised never to copy it word for word or take it out of the room.

When you have that mosaic built from UK sovereign defence, health, roads, power networks, power stations, and our major industrial bases, you have a detailed understanding of virtually every aspect of the sovereign United Kingdom. For an adversary, or even a nation with whom we have a special relationship, that picture is worth more than all the fine art on Earth.
Click to expand...

A second source, whom I gather from the description(s) is also a MoD systems engineer with an intelligence background backs her/him up.

Intelligence source said:
(Palantir probably has) ... a complete profile on the whole UK population. They have visibility into wildly different focus areas, yet their data is all condensed into one foreign supplier’s control/visibility. At the very least I’d call that a security risk.
Click to expand...

Other quotes, apparently from them but not individually attributed.

Both MoD sources said:
Palantir does not need to own the data or even have stewardship. They can extract, transform and exploit the metadata to build their own rich picture.

...

(On metadata and restricted information) A parcel is sent out by a defence supplier with a Nato part number, an address and an arrival date. Even if the label is a QR code and isn’t human-readable, the data it contains would allow Palantir to know that a nuclear submarine would be in Diego Garcia on 4 April. Those three bits of information – the part number, Diego Garcia and 4 April – are, individually, completely unclassified. Together, they are secret.
Click to expand...

From this article:

https://www.thenerve.news/p/palanti...vernment-data-insights-security-state-secrets
The sources claim that the Palantir contracts as a whole pose "a national security threat to the UK" and that that notions about data ownership and sovereignty in relation to the company's access on the part of the UK are "ignorant".

Somewhat akin to how private citizens are now unthinkingly giving "AI agents" (and more significantly, the companies providing them) truly unprecedented access to their own data, apparently not realizing how fundamentally they're compromising their privacy in the process compared to all precedents. For example, I've seen first hand accounts on top people from premier academic institutions using these agents for the most mundane reasons and giving blanket access to all their research in the process; ostensibly some of the most accomplished, analytical minds around FOMOing their way around in this brave new LLM environment.
 
Last edited:
Reddington777 said:
https://breakingdefense.com/2026/06...und-ai-tools-ill-suited-to-tactical-planning/

Im sure something many of you will be screaming "I told you so" for
Click to expand...

Within the military (and everywhere else) LLMs are being used to solve problems that these tools are not suited to.

At the same time there are certain problems that “AI” is really, really well suited for. Some of these problems are opportunities for small companies to become very big companies overnight.
 
quellish said:
Within the military (and everywhere else) LLMs are being used to solve problems that these tools are not suited to.
Click to expand...
Yeah absolutely.

I think the closer you get to the ground / infantry combat, the harder it is for LLMs to work well. The article talks specifically about AI struggling with 3d space in mission planning for ground troops, which is a much more messy, complex environment than the sky where many boundaries are human defined rather than physical. Its also not the same as mission planning for PGMs either, where the blast itself can easily erase obstacles and hit the target as long as you've got some 3d coords.

That extra precision that may be required for ground ops possibly requires some computer vision or some form of 3d reconstruction in order to improve mission planning.
 

New AI ‘Agent Network’ Could Gather Intel Faster for Strike Packages​

https://www.airandspaceforces.com/new-ai-agent-network-intel-strike-packages/
The Maven Smart System, which grew from the department’s Project Maven more than half a decade ago, uses AI to sort through data streams and identify potential targets. It played a key role in the early stages of Operation Epic Fury against Iran, when, according to reports, operators used the system to craft strike packages for more than 1,000 targets. Commander demand for the system is growing.
Click to expand...
 
Forest Green said:

New AI ‘Agent Network’ Could Gather Intel Faster for Strike Packages​

https://www.airandspaceforces.com/new-ai-agent-network-intel-strike-packages/
Click to expand...

New? We've been reading about IDF real-time target generation using AI, UAVs, Unit 8200, etc. for years. Their "Server in the Sky" (SITS) is fascinating: https://archive.ph/lJtqr

"New" used as a promotional device is not so much troubling as the attempt to pretend there is no connection. (SUS:oops:)

The opening sentence: "A new Pentagon initiative could shrink the timelines for air and space commanders to gather and sort through the intelligence necessary to conduct operations."

I've personally been reading about "closing sensor-shooter (operational) loops" for the past 5 years. :rolleyes:
 
Last edited:
VishnuOsiris said:
New? We've been reading about IDF real-time target generation using AI, UAVs, Unit 8200, etc. for years. Their "Server in the Sky" (SITS) is fascinating: https://archive.ph/lJtqr

"New" used as a promotional device is not so much troubling as the attempt to pretend there is no connection. (SUS:oops:)

The opening sentence: "A new Pentagon initiative could shrink the timelines for air and space commanders to gather and sort through the intelligence necessary to conduct operations."

I've personally been reading about "closing sensor-shooter (operational) loops" for the past 5 years. :rolleyes:
Click to expand...
Presumably the system in question is new, even if what it's doing is not and perhaps it does it better maybe??
 

Lockheed Martin and U.S. Navy Demonstrate AI-Powered Submarine Hunter at RIMPAC 2026​

https://www.navalnews.com/naval-new...e-ai-powered-submarine-hunter-at-rimpac-2026/
This summer at RIMPAC 26, Lockheed Martin collaborated with the U.S. Navy to successfully demonstrate a rapidly retrainable AI/ML sonar expert for antisubmarine warfare (ASW). The system, known as SensorMAX™, monitors spectral energy including subsurface acoustic data to assist the crew with rapid threat detection. Augmented by over-the-air (OTA) model update capability, SensorMAX aims to dramatically improve the U.S. Navy’s subsurface detection capabilities.
Click to expand...
SensorMAX uses AI/ML technology to understand ocean sounds. Taking in subsurface acoustic data from a network of sensors, SensorMAX provides a continuous data stream to the aircrew and sends information to the ground station. When a new target is identified, system operators label the identified noise signature, retrain the model on the new data and push a small update packet back to the aircrew.
Click to expand...
This closed-loop workflow – capture, label, retrain and deploy – enables the crew with continuously enhanced situational awareness throughout their mission. Retraining the model takes under 5 minutes per iteration.
Click to expand...

Mission Impact​

The success of the exercise proves that an integrated force of MH-60R helicopters and AI-enabled acoustic experts can detect and track underwater threats faster and more effectively than traditional methods, accelerating the Navy’s vision for fleetwide adoption of crew-uncrewed teaming.
Click to expand...

Core Capabilities​

  • Cross-sensor AI training. Lockheed Martin’s SensorMAX can process data from a variety of sensors supporting operations with commercial and defense applications.
  • Works everywhere. Because SensorMAX is lightweight and portable, the same “brain” could be used for data processing in the air, at sea or on land.
  • Rapid inflight classifier update. The ability to update the AI classifier during flight shortens detection-to-engagement time when new threats are detected, giving the helicopter crew a decisive edge against enemy submarines.
  • Model updates OTA. Model updates are delivered over-the-air. Lockheed Martin’s collaboration with FUSE Inc. enables secure, resilient software delivery to deployed aircraft, allowing SensorMAX’s trained AI models and mission software to be updated rapidly without interrupting operations.
Click to expand...
By demonstrating platform-agnostic AI intelligence and classification, the future fleet can leverage any acoustic sensor to expand the protective ring around high-value assets and accelerating the ASW kill chain.
Click to expand...
 
https://twitter.com/x/status/2098770468081734081
View: https://x.com/Drone_Wars_/status/2098770468081734081?s=20

The U.S. Army has moved the TITAN program from prototyping into production. Palantir and Anduril have received a combined $192 million — $127 million for Palantir and $65 million for Anduril — to build the first production batch of these mobile intelligence nodes, designed to speed the collection, fusion, and delivery of targeting data against distant threats.TITAN’s purpose is to pull sensor data from space, air, and ground sources into a single, AI-enabled system and get usable targeting information to units faster. It is not just another comms vehicle. It is meant to shorten the path from detection to decision — and from sensor to shooter.
Click to expand...
Palantir leads overall architecture, software, and production. Anduril supplies the ruggedized vehicle-mounted hardware, including power, thermal management, cooling, and mission shelters. Two variants are planned: a heavier Advanced version for tactical trucks (FMTV) at corps and theater level, and a more mobile Basic version for JLTVs (with additional work on Infantry Squad Vehicles).The first four Advanced and four Basic systems are scheduled for delivery within 18 months. They will join the nine prototype units the Army already keeps in operational use.
Click to expand...

https://twitter.com/x/status/2098830248725119426
View: https://x.com/outerspacetoday/status/2098830248725119426?s=20


NEW: The Pentagon wants AI to turn raw space and missile-warning data into actionable threat intelligence in as little as 2 seconds.A Defense Innovation Unit solicitation released on Sept. 10 seeks a “Space Threat Intelligence Synthesis Engine” that can ingest live video, satellite imagery, radar, geospatial data, sensor feeds, and classified intelligence, then identify threats and generate confidence-scored alerts almost instantly.
Click to expand...
The scale is aggressive, with up to 5 GB of incoming data per minute, a maximum 5-second data-to-display latency, and a 2-second objective. The system would distinguish closely spaced objects, attribute events to specific threat actors or platforms, and continuously update its understanding of weapons-system capabilities.
Click to expand...
The Pentagon wants machine-to-machine APIs that feed intelligence directly into automated command-and-control systems, effectively compressing the time between detecting a potential threat in space and acting on that information. The bids are due Sept. 24.
Click to expand...
 
Last edited:

Similar threads

Back
Top Bottom