The AI Safety Menu

All laws

Humans control nukes

Human control of nuclear launch decisions. Only humans may decide to launch nuclear weapons.

Lead costAlmost noneof America's lead over China, over 3 years
Cuts p(doom) by~0.08%from a 5% starting estimate
Holds up○ Training○ Lab's own use○ Public release
Enact it on the menuSee the findings

What it does

Only people, not AI systems, can decide to use nuclear weapons. The U.S. and China affirmed this in 2024, and U.S. law states it as policy.

Applies to: U.S. nuclear command and control.

Where things stand

U.S. policy since the FY2025 defense bill. A stricter funding ban has not been enacted.

Why it costs almost no lead

It governs the military, not AI labs, so it costs America's AI lead nothing.

Biggest unknown: None for the lead.

Why it lowers p(doom) by ~0.08%

Keeps AI from ever deciding on its own to launch nuclear weapons.

It keeps a human decision on nuclear launch. That addresses a narrow but severe risk.

The strongest case that it costs more

There is no plausible case that it slows frontier AI. The debate is whether a statement of policy is strong enough.

The debate

For

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.