Automated military escalation becomes a global war

The scenario

Scenario, pathway and assumptions based onA Taxonomy of Omnicidal Futures Involving Artificial Intelligence ↗

The scenario begins with countries becoming less dependent on one another. Robotic mining, manufacturing and agriculture reduce their need for foreign labor and goods, while domestic energy production makes trade less essential. Leaders become more willing to risk diplomatic relationships they previously needed. Military preparation grows alongside this economic independence.

Governments replace human soldiers with large drone forces, initially presenting this as a way to fight more cheaply and with fewer military casualties. Human operators gradually surrender control because AI can coordinate enormous fleets and respond faster. During confrontations, leaders demonstrate resolve by programming territorial boundaries and automatic retaliation directly into their forces. What once required another political decision can now happen as soon as a condition is met.

A boundary violation triggers a counterattack that crosses the opponent’s next threshold. Its response then activates another rule, producing a widening chain of attacks. Leaders trying to stop the process face both the practical difficulty of interrupting automated operations and the domestic political cost of appearing to back down. The systems continue executing commitments made earlier, even as the situation changes around them.

In Critch and Tsimerman’s ending, these conflicts spread into a worldwide war. Machines keep fighting even where the human populations they served are dying out, and almost everyone is eventually killed. The authors do not describe how every region is reached or calculate the war’s ultimate death toll. The existential endpoint therefore requires an additional step beyond rapid escalation: sustained global destruction without a successful interruption.

How it could unfold

1. Economic independence weakens restraints on confrontation.

2. Leaders delegate force and encode automatic retaliation.

3. Opposing response rules drive a widening conflict.

4. War persists and spreads despite catastrophic human losses.

Evidence and its limits

The ICRC warned in July 2022 that autonomous weapons could accelerate force beyond human control and make escalation unpredictable. That supports concern about the proposed feedback mechanism; it does not establish Critch and Tsimerman’s global-war or near-extinction outcome.What you need to know about autonomous weapons ↗

What this depends on

Countries must give automated forces enough discretion to escalate, while retaining too little practical or political capacity to interrupt them. The ending also assumes fighting remains sustainable as societies collapse, rather than stopping through breakdown, containment or negotiated limits.

Objections and barriers

ICRC and SIPRI identify restrictions on targets and operating environments, together with human supervision, as practical ways to retain control. Our assessment is that effective intervention would break the scenario’s automatic escalation chain; the paper does not show that these controls must fail everywhere.Limits on Autonomy in Weapon Systems ↗A Taxonomy of Omnicidal Futures Involving Artificial Intelligence ↗