What a slowdown looks like.

1. No superintelligence until it’s safe. No smarter-than-human AI without scientific consensus and public consent.
2. Binding rules for frontier labs. Mandatory testing, transparency, and liability. Voluntary pledges have failed.
3. Protect people first. Guardrails on AI companions and kids’ products; a plan for displaced workers before they’re displaced.
4. A deal, not a race. Verified international agreements, the way we handled nuclear weapons.