The AI Safety Menu

All laws

Whistleblower protection

Whistleblower protections. Employees can report catastrophic-risk concerns to regulators without retaliation.

Lead costAlmost noneof America's lead over China, over 3 years
Cuts p(doom) by~0.1%from a 5% starting estimate
Holds up○ Training○ Lab's own use○ Public release
Enact it on the menuSee the findings

What it does

Lab employees can tell regulators about serious safety risks without being fired or sued. Large labs also have to run an anonymous internal reporting line.

Applies to: Frontier developers.

Where things stand

Anthropic and OpenAI both publicly dropped non-disparagement clauses that could have chilled safety disclosures, and both describe internal reporting channels in their frameworks. The law makes that permanent.

Why it costs almost no lead

It changes who can talk to whom, not how models get built. Running the reporting line displaces a tiny amount of research: under a day.

Biggest unknown: None. This item is on the menu to make a point about what regulation mostly consists of.

Why it lowers p(doom) by ~0.1%

Lets employees warn regulators about hidden dangers without being fired or sued.

Insiders are often the first to see problems, and several past warnings about lab safety came from employees. Protecting them is one of the cheaper safeguards.

The strongest case that it costs more

There is no serious case for a lead cost. The strongest argument against this provision is about litigation risk and morale, not about the race.

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.