Lethal Autonomy

Frank Kendall, who served as US Air Force Secretary from 2021 to 2025 and earlier as the Pentagon’s acquisition chief, writes in Lethal Autonomy from the vantage point of someone who has spent decades deciding what the American military buys. Published by Knox Press in July 2026, the book is a domain-by-domain survey of how artificial intelligence and autonomous weaponry are reshaping conventional warfare.

Kendall’s central argument is that lethal autonomy will render current militaries almost entirely obsolete. The drone saturation seen in the Russia-Ukraine war, in his framing, is only the opening indication. He sets out the attributes that have historically driven military-technical change and the analytical basis for assessing new operational concepts, then applies that framework across the traditional domains — ground, sea surface, subsurface, and air — and extends it into space, cyberspace, and multi-domain operations.

The book’s space chapter illustrates the method. Kendall splits space into two conflicts: an offensive campaign to eliminate an adversary’s space capabilities, and a defensive campaign to protect one’s own. He argues the United States is behind China in fielded space order of battle, particularly in numbers of ISR satellites and anti-satellite systems, and recounts the 2014 briefing he gave the Deputy’s Management Action Group laying out three choices: keep building exquisite multibillion-dollar satellites and try to defend them, redesign the architecture for resilience, or reduce reliance on space entirely. He describes the Third Offset Strategy work under Deputy Secretary Bob Work, the Long-Range Research and Development Planning study chaired by Stephen Welby, and the later contributions of General John “Jay” Raymond as the first Chief of Space Operations, the Space Warfighting Analysis Center under Andrew Cox, and successors Chance Saltzman and Troy Meink. He treats space less as “high ground” than as a no man’s land: a chessboard on which both sides preposition pieces before the game starts, and where the rewards for going first make the domain tactically and strategically unstable in a period of tension. Kinetic, directed-energy, cyber, and electronic attacks blend together, and because orbital mechanics compress decision time, battle management becomes a problem for automated systems, with humans monitoring and overriding rather than directing in real time.

The other chapters give ground, sea surface, subsurface, air and cyberspace the same treatment: the concepts, building blocks and complexities that have to be weighed in designing forces that incorporate lethal autonomy. Lethal Autonomy sits apart from the broader AI-and-war literature in two ways: it is written by an acquisition insider rather than a strategist or ethicist, and it treats autonomy as an operational and force-design problem rather than a moral one. Readers looking for a policy-level argument about killer robots will find a different book; readers interested in how a former service secretary thinks the US military’s order of battle has to change will find Kendall’s case laid out in detail.