CA Frontier AI Safety — Requirements, Penalties & Compliance Guide

Summary

Requires frontier AI developers to implement safety protocols, incident reporting, and compute tracking.

Who it applies to

Scope: frontier ai. This regulation applies to deployers and/or developers of AI systems in the frontier ai category operating in or affecting CA.

Penalties

Up to $1M per violation. Enforced by Attorney General.

Key requirements

This regulation has 6 distinct requirements. Deep-dive guides with step-by-step implementation, evidence needed, and passing criteria:

• Safety Protocols
• Incident Reporting
• Compute Tracking
• Model Inventory
• Red-Team Testing
• Documentation

Effective date

Jan 1, 2026. Currently in effect.