The AI Safety Menu

All laws

Pause with China

Verified six-month pause with China. Both countries pause frontier training for six months, with verification.

Cuts lead by~11 daysof America's lead over China, over 3 years
Cuts p(doom) by~0.4%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. and China both stop training their most advanced models for six months, with inspections and chip tracking to check the other side is complying.

Applies to: Frontier training runs in both the U.S. and China.

Where things stand

There is no U.S.–China agreement on frontier AI. In November 2024 Biden and Xi affirmed human control over nuclear launch decisions. In September 2026 the U.S. proposed an AI dialogue with a mechanism for notifying each other of serious incidents.

Why it costs ~11 days

Both countries stop, so most of the cost cancels. Verification tools aren't ready, so we assume China keeps some progress through cheating or hard-to-check work. Net: about 11 days a year in.

Biggest unknown: Whether a pause can be verified. The key technologies have not been built or tested.

Why it lowers p(doom) by ~0.4%

Stops both America and China from racing ahead while safety catches up.

Both sides stopping removes the race pressure behind risky shortcuts. It's temporary and hard to verify, so the lasting effect is moderate.

The strongest case that it costs more

A verification regime would take years to build, so a pause agreed quickly would rely on trust. If China kept most of its progress, the U.S. would lose close to the full six months.

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.