Agent AI Army

EDITORIAL / POLICY ANALYSIS

Autonomous weapons are already a policy question. Most coverage still treats them like science fiction.

Agent AI Army reads the public record on military AI — doctrine, procurement, ethics debates — and writes what's actually there, sourced and dated.

Get new analysis by email

One or two pieces a month. No noise.

WHAT WE COVER

Three questions we keep returning to.

01

Doctrine & policy

How militaries and governments are writing autonomy into official doctrine — and where the language is still vague on purpose.

02

The human-in-the-loop debate

What "meaningful human control" actually means across different systems, and where it quietly disappears.

03

Reading the public record

Procurement documents, think-tank reports, official statements — read closely instead of summarized from a press release.

DEFINITIONS

What counts as an autonomous weapon?

A lethal autonomous weapon system (LAWS) is a weapon that, once activated, can select and engage targets without further human intervention. That is the working definition behind the UN debates — and it is harder to apply than it sounds, because modern air-defence systems and loitering munitions already automate parts of that chain. The policy argument is rarely about whether a system is "autonomous"; it is about which decisions stay with a person.

IN THE LOOP

The system finds and proposes targets; a person approves every engagement before it happens.

ON THE LOOP

The system engages on its own; a person supervises and can abort — if there is time.

OUT OF THE LOOP

The system completes the entire targeting cycle with no real-time human role.

Most of today's debate lives in the middle row: supervision that exists on paper but shrinks in practice as engagement windows get shorter. Primary sources worth reading directly: the UN's CCW process and the ICRC position on autonomous weapon systems.

EDITORIAL RULES

What this site is, and what it deliberately isn't.

Agent AI Army is a publication. We read and analyze information that's already public — official reports, procurement filings, academic and think-tank research, policy statements. We don't publish technical specifications, classified material, or anything not already available through open sources.

We don't offer consulting, integration, or technical services to any government, contractor, or private buyer, and we don't broker introductions between them. If that's what you're looking for, this isn't the right site.

Publisher, not vendor. That line doesn't move.

Editorial disclaimer

This site provides commentary and analysis on publicly available information. It does not offer technical, advisory, or brokerage services related to weapons or defense systems, and does not represent the views of any government or military organization.

QUESTIONS

A few things worth clearing up first.

Is this site affiliated with a defense contractor or government agency?

No. Agent AI Army is an independent publication. We have no contracts, sponsorships, or advisory relationships with defense contractors, militaries, or government agencies.

Do you cover classified or leaked material?

No. Every analysis is built exclusively from information already in the public domain — official publications, open research, and public statements. We don't seek out or publish leaked or classified content.

Are autonomous weapons banned under international law?

No dedicated treaty bans them. They fall under existing international humanitarian law, and states have discussed possible new rules for years within the UN CCW expert group — so far without a binding instrument. Tracking that gap between debate and law is much of what we do.

What does "meaningful human control" actually require?

There is no agreed legal definition — which is precisely why the phrase matters. In the policy debate it stands for a human having enough understanding, time and authority to make a real decision about force, rather than approving whatever the system proposes. Where that threshold sits is the core dispute.

STAY INFORMED

New analysis, roughly monthly.

Leave your email for new pieces as they publish. Nothing else.

No spam, no third-party sharing. Unsubscribe anytime.