The Download: how AI is used for military targeting, and the Pentagon’s war on Claude

2026-03-13 · Source: MIT Technology Review · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Cybersecurity & Data Privacy · Depth: Fundamental Awareness, short

Summary

The US military is exploring the use of generative AI systems, such as OpenAI's ChatGPT and xAI's Grok, to prioritize and recommend targets for military strikes, according to a Defense Department official. This process would involve feeding a list of potential targets into a classified AI system, which humans would then prompt to analyze and rank them, retaining responsibility for evaluating the AI's recommendations. Concurrently, the Pentagon's CTO has expressed concerns that Anthropic's Claude model could "pollute" the defense supply chain due to its baked-in "policy preference." This development occurs as Ukraine offers its battlefield data for AI training to allies, and Meta postpones an AI launch due to performance issues against rivals.

Key takeaway

For CTOs and VPs of Engineering evaluating AI integration into sensitive operations, you should critically assess the policy preferences and training data of generative AI models. The Pentagon's concerns about Claude highlight that inherent model biases can be a significant barrier to adoption in defense. Prioritize models with transparent architectures and verifiable neutrality, and ensure robust human-in-the-loop validation processes are in place to mitigate risks associated with AI-driven recommendations.

Key insights

Generative AI is being explored for military targeting decisions, raising questions about AI policy and data utility.

Principles

• Human oversight is critical in AI-driven targeting.
• AI model policy preferences impact defense integration.

Method

A generative AI system receives target lists, analyzes information, and prioritizes targets, with human operators verifying and evaluating the recommendations before action.

In practice

• Assess AI model biases for defense applications.
• Explore AI for data analysis in high-stakes contexts.

Topics

• Military AI
• Generative AI
• AI Policy
• Large Language Models
• Battlefield Data

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Tech Journalist, Policy Maker, General Interest

Related on AIssential

• Pentagon Sees Bigger Role for AI in Setting Military Targets — The Pentagon intends to significantly expand AI's role in military target identification and selection.
• Anthropic and the Pentagon are reportedly arguing over Claude usage — AI companies face pressure from the U.S. military to permit broad use of their models, sparking policy disputes.
• Key Questions on the Role of Technology in the Expanding Middle East War — AI's inherent flaws, information environment degradation, and warfare privatization are critical issues in modern conflict.
• US military leans into AI for attack on Iran, but the tech doesn’t lessen the need for human judgment in war — Effective military AI use depends more on organizational factors and human judgment than on technology's inherent capabilities.
• SUNY Poly defense contract: Using AI, new tech to help army strategize - Utica Observer Dispatch — Integrating high-velocity AI into army decision-making frameworks provides a strategic advantage for multi-domain conflict preparedness.
• Where OpenAI’s technology could show up in Iran — OpenAI's military integration highlights the dual-use dilemma of advanced AI in conflict zones.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT Technology Review.