Palantir showed how the military can use AI chatbots to plan operations

Palantir Technologies has conducted a series of closed demonstrations showing how large language models — specifically Claude from Anthropic — can be used by the military to analyze intelligence data and generate proposals for operational planning. Materials from these demonstrations and Pentagon documents examined by Wired journalists provide the first public view of how AI chatbots are being integrated into military command systems. Palantir — a company specializing in data analytics for government agencies and intelligence services — has long worked with the U.

S. Department of Defense. However, public demonstrations of chatbot interfaces specifically for military planning have appeared for the first time.

During the demonstrations, the AI system received intelligence data, analyzed it, and offered officers concrete next steps — ranging from logistical solutions to tactical recommendations. A key detail: in some of the demonstrations, Claude from Anthropic was used as the language model — a company that positions itself as a priority developer of safe AI. This creates a complex context.

Anthropic publicly emphasizes responsible use of technology, yet its developments are proving to be at the foundation of military planning systems. Anthropic is not the first civilian AI vendor to collaborate with the defense department, but the nature and scale of integration have sparked a new wave of discussions in the industry. Pentagon documents confirm: this is not about pilot experiments on paper, but about actively developing real combat applications.

LLM-based systems are being considered as tools to accelerate the so-called OODA cycle — Observe, Orient, Decide, Act. In military operations, the speed of processing intelligence data and making tactical decisions is critically important, and AI assistants can provide tangible advantage here. Palantir positions its AIP platform (Artificial Intelligence Platform) as a solution for government customers with emphasis on data controllability and security.

During the demonstrations, the operator asked the system questions in natural language, and it aggregated data from multiple sources and generated a structured response with recommendations. The final decision, according to the company's official position, always remains with the human. However, critics point to risks.

When an AI system suggests next steps in a combat situation, the line between recommendation and decision becomes blurred. Cognitive research shows: people tend to trust automated systems even with incorrect conclusions — especially under the stress and high information load characteristic of military commanders. This is a serious problem that the industry has yet to solve.

A separate issue is data. Intelligence information processed by such systems falls into the highest secrecy categories. Palantir has the necessary clearances and years of experience working with classified infrastructure, however the integration of commercial LLMs raises new questions about security, model control, and potential vulnerabilities.

What is happening is part of a broader trend. The militaries of leading powers are actively exploring AI capabilities to accelerate command and control. The gap between public declarations about responsible AI and real-world application of technology in military systems is becoming increasingly evident.

For the industry, this is a signal: companies creating large language models inevitably become part of the defense technology stack — regardless of stated values.