The AI Safety Menu

All laws

Killer robots, no limits

Autonomous weapons without human control. The U.S. builds and deploys lethal autonomous weapons with no requirement for human control.

Lead costAlmost noneof America's lead over China, over 3 years
Raises p(doom) by~0.15%from a 5% starting estimate
Holds up○ Training○ Lab's own use○ Public release
Enact it on the menuSee the findings

What it does

The U.S. would field weapons that can pick and strike targets on their own, without a firm rule that a human must meaningfully control each use of force. It would also keep blocking a binding treaty.

Applies to: Military weapons programs.

Where things stand

DoD Directive 3000.09 (updated Jan 2023) requires 'appropriate levels of human judgment over the use of force' and senior review before fielding, but does not ban autonomous weapons. The U.S. says human control is not a legal requirement as such.

Why it costs almost no lead

It changes weapons, not frontier AI. Under a day gained.

Biggest unknown: What the revised 3000.09 actually says about human judgment, and whether the Nov 2026 CCW Review Conference opens treaty talks without the U.S.

Why it raises p(doom) by ~0.15%

Not a safety law. Lets weapons choose and attack targets without a human deciding.

Weapons that select targets on their own add paths to accidental escalation and lower the bar for war.

The strongest case that it costs more

Supporters argue autonomous systems can be more precise than humans, and that China will build them regardless.

The debate

For

Against

Sources

Rough starting points, not precise forecasts. Lead costs assume China doesn't depend on U.S. models, the case least favorable to safety laws, and count 3 years. On the menu you can change every assumption and put in your own numbers. Last priced 2026-09-26.