Who answers when a machine decides?
Autonomous weapons, AI-assisted targeting and autonomous cyber agents each promise speed and precision. Each also strains the oldest rule of war: someone must answer for the use of force.
This book follows the
accountability gap through three dimensions:
technical (can the system be tested and its decisions reconstructed?),
legal (what is binding law, and what is only a political promise?) and
organisational (who really decides, and who answers?).
Inside this edition
- What the 2026 UN expert process under the CCW did and did not agree, separating binding law, political commitments and proposed rules
- Why human oversight often fails in practice: automation bias, time pressure and the "moral crumple zone"
- Drones from Predator to Kızılelma, decision-support targeting, and the reported ECHOTRIBBLE autonomous cyber-agent programme
- How a few companies and data centres now shape military AI
- Satellite imagery, wildfire AI and the surveillance problem
- Testing and red teaming of unmanned and autonomous systems
- Nuclear command, and the position of small states in the Western Balkan
Every claim is labelled:
established fact,
probable or
unconfirmed, and conclusions are never stated more strongly than their weakest evidence.
Written for policy analysts, lawyers, students, military professionals and informed readers.
An analytical work based on public sources, current to 15 October 2026. It contains no classified information or operational detail.