Palantir Demonstrations Illustrate Potential Military Applications of AI Chatbots for Creating War Strategies

When a user inquires, “What enemy military unit is in the area?” the AIP Assistant infers that it’s “likely an armor attack battalion based on the equipment pattern.” This leads the analyst to request an MQ-9 Reaper drone to survey the situation. They subsequently ask the AIP Assistant to “generate three courses of action to target this enemy equipment,” and within moments, the assistant proposes attacking the unit using either an “air asset,” “long-range artillery,” or a “tactical team.” The user instructs the assistant to relay these options to a fictitious commander, who ultimately opts for the tactical team.

The concluding steps occur swiftly: The analyst asks the AIP Assistant to “analyze the battlefield,” followed by a request to “generate a route” for troops to reach the enemy, and finally to “assign jammers” to disrupt their communications equipment. Within seconds, the analyst conducts a final review of the battle plan and gives the order for troop mobilization.

In this context, Claude would serve as the “voice” of the AIP Assistant, as well as the “reasoning” behind the generated responses. Other AIP demonstrations exhibit similar interactions with large language models. For instance, a blog published last week by Palantir highlighted how NATO, a Maven Smart Systems client, could utilize an AIP Agent within their tool.

In one visualization, Palantir illustrates how a third-party defense contractor can choose from various built-in AI models offered by Palantir, including different versions of OpenAI’s ChatGPT and Meta’s Llama. The user selects OpenAI’s GPT 4.1, although it appears that a soldier might also have the option to select Claude.

An analyst then examines a digital map displaying the locations of troops and armaments. In a panel titled “COA” (courses of action), they click a button that activates a tool powered by GPT-4.1 to produce five potential military strategies, one of which is called “Support-by-Fire-then-Penetration-Shock-and-Destruction.”

Another instance demonstrates how the system can be utilized to interpret satellite images: The analyst identifies three tanker truck detections on a map, loads them into the AIP Agent’s chat interface, and requests it to “interpret” the imagery and propose options for action.

Claude may also assist the military in generating intelligence assessments that could influence future strike planning. In June 2025, WIRED observed a presentation by Kunaal Sharma, a public sector lead at Anthropic, showcasing how the enterprise version of Claude could be employed to create “advanced” reports about a real Ukrainian drone strike known as “Operation Spider’s Web.” In the demonstration, Sharma noted that Claude was solely utilizing publicly available information. However, he indicated that by collaborating with Palantir, the federal government could also leverage internal datasets.

“This is typically something that I might spend five hours with a cup of coffee, researching on Google, delving into think tanks, and then writing reports and citations, among other tasks,” Sharma mentioned. “But I don’t have that kind of time.”

In the demo, Sharma requested Claude to develop an “interactive dashboard” featuring information about Operation Spider’s Web and then translate it into “object types” for analysis in Foundry, one of Palantir’s standard software products. He also asked Claude to craft a detailed analysis of recent occurrences in Russia’s border regions, as well as a 200-word summary of the operation’s “military and political effects.”

“Honestly, I’ve been reading these types of documents for twenty years—I used to write them, and I was an academic myself,” Sharma stated. “This is actually quite impressive.”