How this scenario was built

This is a scenario, not a forecast

The project follows three regulatory choices through procurement, conflict, institutional incentives and crisis conditions to expose consequences that static analysis misses. No probabilities are assigned to the paths, and none should be inferred. The value of a scenario lies in the visibility of its reasoning, not the authority of its conclusions.

2040 is an institutional horizon, not the expected date of first harm

The dangerous dynamics in this scenario — compressed decision cycles, routine target-recognition assistance, terminal autonomy at scale — are documented in the 2020s and accelerate through 2027–2033. The year 2040 is kept as the comparison point because institutions, precedents and industrial structures take that long to harden. Reading the site as predicting first harm in 2040 would invert its argument.

Evidence basis, corroboration and confidence are labelled separately

Every chapter states its basis: documented (reliable evidence), reported (a credible but not fully verified claim), projected (a causal proposition extrapolated from identified incentives) or speculative (lower-confidence, included for consequence). A projection is not evidence and is never described as such.

Reported and documented material additionally carries a corroboration status — corroborated, partly corroborated or uncorroborated. The reported 2024 Ukraine 'killer drones' claim is uncorroborated: a single named industry source, with no operational logs, official confirmation, geolocated footage or independent investigation in the public record. It is used only as an attributed claim and an exercise in evidentiary uncertainty.

Current factual claims have a July 2026 evidence cut-off.

The common stress tests are comparative devices

All three paths face the same four classes of external pressure — tempo (2027), civilian harm (2028–2029), saturation (2031) and escalation (2033–2034). The circumstances differ between paths; the pressure is comparable. This is a deliberate device to stop the branches becoming three unrelated novels: the reader can hold the pressure constant and watch the governance architecture vary.

Scenario indices are illustrative

The monitor's indicators are authored editorial judgements on a rounded 0–100 scale, split into pressure gauges (higher = greater exposure) and safeguard gauges (higher = stronger). They exist to make trend direction legible. No indicator value is a statistic, and none should be quoted as one.

The scenario has a normative point of view

It argues that accountability — the technical capacity to reconstruct what a system did, the institutional capacity to determine whether it was lawful, and the legal capacity to assign consequences — is part of defence capability, and that lawfulness and civilian protection are requirements in their own right. That position shapes the writing and deserves the same scrutiny the scenario applies to the policies it examines. Each path is therefore written with real, durable benefits and real, structural costs: Broad Restriction achieves lasting prohibitions and norms; Tiered Assurance pays for its accountability in money, speed, staffing and forgone strikes, and still buries civilians in 2029 and 2037.

Frontier AI is a conditional accelerant, not a premise

The scenario's dangers run on narrow, current technology: recognition, navigation, ranking, interception. Highly capable general models matter conditionally — when connected to intelligence pipelines, cyber access, command systems, industrial capacity or physical weapons. Those connections are decisions, and governance can reach decisions. AGI is not the assumed engine of any event in this timeline.

Legal content: proposals, hypotheticals and required review

The accountability section proposes reforms — a corporate failure-to-prevent offence, individual offences for knowing or reckless assistance, regulatory offences, a continuing due-diligence duty, and an International Autonomous Weapons Agency. These are policy recommendations, not statements of current law, and the site labels them as such. Certification is treated throughout as evidence of compliance, not immunity.

The site distinguishes: a harmful system outcome; an IHL violation by a conflict party; a regulatory or contractual breach; corporate civil or administrative liability; and individual criminal complicity in a war crime. The ICC prosecutes natural persons, not corporations; corporate criminal liability must ordinarily be pursued under national law. Criminal liability requires the elements of the offence, including the required mental element and a sufficient contribution — a bad product, public controversy or employment at a supplier is not enough by itself. State, command, operator and supplier responsibility may coexist; none cancels another.

The supplier case study is a legal hypothetical built on one narrow, sourced fact (a publicly reported defence partnership). No knowledge, operational use or offence is asserted as fact, and the named example can be replaced by a neutral 'Company P' in a single content field. All legal content requires review by qualified UK and international-criminal-law counsel before publication.

Planned review

Before formal use, this project should be reviewed by international humanitarian law specialists; serving or former operators with targeting experience; procurement and contracting specialists; AI-assurance and human-factors researchers; defence-industry practitioners; and civil-society researchers on autonomous weapons. Disagreements from that review should be published alongside the scenario, not resolved silently.