Exclusive eBook: How AI is becoming the next military advisor
TL;DR
MIT Technology Review packages six James O’Donnell stories as a subscriber-only eBook on how militaries are using AI models as decision aids. The stories were originally published between April 11, 2025, and April 21, 2026, and the publisher says they were updated to reflect recent developments. The focus is less on sci-fi killer robots and more on models that help commanders, analysts, and planners assess situations, prioritize information, and make calls.
Nauti's Take
The real story is not the armed robot, it is the model in the briefing room. If you build AI for these workflows, you are not making neutral assistance, you are shaping command chains, priorities, and risk calls.
Pretty accuracy folklore is not enough here.
Briefingshow
Military AI becomes dangerous when models do more than sort data and start shaping how humans assess targets, risks, and escalation. A human approval click is not enough if the options were already framed by a system nobody fully trusts. The hard question is who understands the model’s limits, spots false confidence, and owns the consequences when advice turns into force.
Sources