The AI Safety Menu

All laws

Chatbot safety rules

Safety rules for AI companion chatbots. AI disclosure, crisis protocols and protections for minors in companion chatbots.

Lead costAlmost noneof America's lead over China, over 3 years
No p(doom) effectNonefrom a 5% starting estimate
Holds up○ Training○ Lab's own use○ Public release
Enact it on the menuSee the findings

What it does

Chatbots must tell users they are talking to an AI, respond to signs of suicide or self-harm with crisis resources, and for minors, remind them to take breaks and block sexual content.

Applies to: Operators of companion chatbots; possibly general-purpose assistants.

Where things stand

California, New York and at least 13 other states enacted chatbot safety laws in 2025 and 2026. The federal GUARD Act passed committee in April 2026. OpenAI publicly backed California's 2026 law.

Why it costs almost no lead

These rules shape consumer products, not training. The product and compliance work displaces under a day of research.

Biggest unknown: Whether general-purpose assistants like ChatGPT and Claude count as companion chatbots, and how heavy age verification would be under the GUARD Act.

Its effect on p(doom)

Protects kids and vulnerable people from harmful chatbot conversations.

It targets harm to individual users, not catastrophic risk.

The strongest case that it costs more

Strict age verification, as in the GUARD Act, could require identity checks for every user of a general-purpose assistant. That is a large product change, though still not a delay to training.

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.