The AI Safety Menu

All laws

Alignment moonshot

Federal alignment research moonshot. Billions a year of federal funding for research on controlling and understanding AI.

Cuts lead by~6 daysof America's lead over China, over 3 years
Cuts p(doom) by~0.3%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 government would spend a few billion dollars a year, compared with roughly $15 million for the federal AI safety institute today, paying scientists to learn how AI models work inside and how to keep them under human control. The results would be shared with labs and used to test the most powerful models before release.

Applies to: Federal research budgets.

Where things stand

Total federal AI R&D is about $3.3B a year (FY2025 NITRD request), of which safety and alignment is a small, unreported share. The UK's Alignment Project has about 27M pounds (15M government).

Why it costs ~6 days

Public money doesn't slow labs. Some researchers would move from building capabilities to safety work: about 2 days.

Biggest unknown: Whether more money can buy alignment progress quickly when the field is limited by talent and by access to frontier models held inside private labs.

Why it lowers p(doom) by ~0.3%

Pays for a huge public push to learn how to keep powerful AI under control.

Many researchers think the bottleneck is knowing how to control advanced AI at all. More funding speeds that up, though money alone doesn't guarantee breakthroughs.

The strongest case that it costs more

Government research programs move slowly, and the best safety work happens inside labs with access to frontier models.

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.