Anduril is trying to solve a problem that sits at the center of modern warfare: too much battlefield information, moving through too many disconnected systems, too slowly. Its answer is Lattice Mesh, an expansion of the company’s software platform that lets outside hardware and software connect to Anduril’s network and exchange data.
The company’s pitch is not simply about better drones or more advanced sensors. It is about making information move quickly enough that people closer to a threat can understand it and act on it. That idea is gaining weight as drones, cameras and remote sensors create more data than human operators can easily sort on their own.
What Anduril showed at its test site
In late November, MIT Technology Review visited an Anduril weapons test site in the foothills east of San Clemente, California. The site included pieces of the company’s hardware suite, including Sentry, a security tower with cameras and sensors, advanced radars and several drones. One drone on display was Ghost, described as an eerily quiet model.
The demonstration centered on Anduril’s software, Lattice. The scenario was built around an incoming drone that could threaten a base belonging to the US military or its allies. It also matched the type of problem at the center of Anduril’s new partnership with OpenAI.
In the exercise, a truck drove toward the base. Sentry automatically recognized it as a possible threat and marked it on a screen. Lattice then asked the human operator whether to send a Ghost drone to monitor it. After one mouse click, the drone flew autonomously toward the truck, using location data gathered by Sentry and passed along through the software.
When the truck moved behind hills, the Sentry camera lost contact. The surveillance drone had already identified it, so the truck’s location remained visible. A person in the truck then launched a drone, which Lattice labeled as a threat. The system asked whether to send a second attack drone, which flew autonomously and locked onto the threatening drone.
The demonstration stopped before impact because Anduril is not allowed to take down drones at that test site. Still, the point was clear: the whole operation could be handled by one person using a mouse and computer.
Why Lattice Mesh matters
Lattice Mesh is meant to extend this type of coordination beyond Anduril’s own equipment. The company announced that more than 10 companies are now building their hardware into the system, including autonomous submarines and self-driving trucks. Anduril has also released a software development kit to help outside companies connect.
The basic idea is that military personnel operating hardware can publish their own data to the network and subscribe to data feeds from other sensors in a secure environment. On December 3, the Pentagon’s Chief Digital and AI Office awarded Anduril a three-year contract for Mesh.
Anduril’s system is also being linked with Maven, a Palantir-operated program that combines information from sources such as satellites and geolocation data. Anduril and Palantir announced on December 6 that the military will be able to use Maven and Lattice together.
That partnership gives the project broader significance. It points toward a battlefield where information from many systems can be fused, shared and acted on through software rather than passed manually from one operator to another.
The military’s connectivity problem
The source article makes clear that the Pentagon does not already have the level of hardware connectivity many civilians experience with phones, laptops and smart TVs. Different defense systems often do not communicate smoothly. Humans may have to translate between systems made by different manufacturers.
That can create slow, manual workflows. One soldier might rotate a camera to look for a drone threat, then send information to another soldier responsible for the weapon that could respond. The information might travel through a low-tech messenger app comparable to AOL Instant Messenger.
Anduril CEO Brian Schimpf framed the company’s goal as a change in where decisions can happen. He said, "You’re going to need to really empower lower levels to make decisions, to understand what’s going on, and to fight." He added, "That is a different paradigm than today."
Zak Kallenborn, a warfare analyst with the Center for Strategic and International Studies, described the challenge as one of volume. Drones, cameras and remote sensors collect large amounts of information, but finding what matters is difficult. As he put it, "There might be something in there, but there’s so much of it that we can’t just set a human down and to deal with it."
Data becomes a strategic asset
The expansion of AI-connected defense hardware could also create a major military data collection effort. Anduril and Palantir said they plan to compile data collected in Lattice, including highly sensitive classified information, to train AI models.
Anduril described the stakes on December 6 by saying, "Exabytes of defense data, indispensable for AI training and inferencing, are currently evaporating." The company’s partnership with OpenAI, announced on December 4, is also connected to model-building work that could benefit from broader sensor data.
Palantir has separately offered AI tools to help the Pentagon rethink how classified data is categorized and managed. Together, these efforts suggest that the value is not only in collecting information, but in preserving it, structuring it and making it usable for AI systems.
Emelia Probasco, a senior fellow at the Center for Security and Emerging Technology at Georgetown University, wrote that the new factor is the Defense Department’s ability to use data in new ways. She said more data and the ability to process it could support accuracy, precision and faster information processing.
The unresolved question of control
Bringing AI closer to military decision-making has already drawn scrutiny. The source article points to Israel’s use of advanced AI models to process intelligence data and generate lists of targets, and notes that Human Rights Watch wrote that such tools "rely on faulty data and inexact approximations."
Probasco said, "I think we are already on a path to integrating AI, including generative AI, into the realm of decision-making." She also examined Maven Smart System, a military system built in 2023 that lets users access sensor data from diverse sources and apply computer vision algorithms to help soldiers identify and choose military targets.
At the same time, Probasco said that an AI system controlling an entire decision pipeline, possibly without human intervention, "isn’t happening" and that "there are explicit US policies that would prevent it."
Anduril also drew a line around what Mesh is intended to do. A spokesperson said, "The Mesh itself is not prescribing actions or making recommendations for battlefield decisions." Instead, the spokesperson said, it is "surfacing time-sensitive information" for operators to consider.
That distinction is likely to remain central. If Anduril’s system works as planned, AI may not need to make the final decision to change the pace and structure of military action. By moving data faster, connecting more systems and making threats visible sooner, Lattice Mesh could still push AI deeper into the operational fabric of war.