The Pentagon’s Dangerous Adoption of AI

February 25, 2026

TL;DR

The Pentagon has accelerated AI integration by awarding contracts to OpenAI, Anthropic, Google, and xAI.
Anthropic's Claude's involvement in an operation to capture Nicolás Maduro has jeopardized its contract and raised ethical questions.
xAI's Grok, known for scandals and a less constrained positioning, has been integrated into military systems, raising alarms.
Commercial Large Language Models (LLMs) are vulnerable to data poisoning attacks, notably attributed to the Russian-linked Pravda network.
The current approach of using commercial AI models is seen as irresponsible due to their lack of accountability, transparency, and potential for misuse in military contexts.
The author advocates for purpose-built AI models for defense, trained on controlled data and rigorously tested, prioritizing reliability over speed and creativity.
The Pentagon's integration of AI raises questions about government transparency, accountability, and trustworthiness, especially when dealing with private industry's less reputable corners.

Continue reading the original article