The AI Safety Menu

All laws

AI Manhattan Project

A government "Manhattan Project" for AI. A federally funded crash program to build the most advanced AI.

Grows lead by~5 weeksof America's lead over China, over 3 years
Raises p(doom) by~0.2%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 federal government funds and leads a crash program to build the most advanced AI, as a 2024 congressional commission recommended. The 2025 Genesis Mission is a much smaller, science-focused version.

Applies to: Federal AI research and compute.

Where things stand

No crash program exists. The Genesis Mission launched in November 2025 and announced over $5 billion in commitments across 278 projects in July 2026.

Why it gains ~5 weeks

Government money and priority access to power could speed the frontier, but private spending is already huge. A program adding about a tenth more compute would gain about 12 days a year in; a much larger one, more.

Biggest unknown: The program's size, and whether China would respond by racing harder.

Why it raises p(doom) by ~0.2%

Not a safety law. A government push to build the most powerful AI first.

A government race could add risk by speeding things up and sidelining safety. It could also put the most powerful AI under public control. Debated, leaning toward more risk.

The strongest case that it costs more

A crash program could provoke China to race harder and abandon any safety cooperation, erasing the gain. Critics also argue it would concentrate power and sideline safety.

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.