Training and controls to ensure models follow human intent and avoid harmful behavior. ← Mainnet Layer 1 (L1) →