Artificial General Intelligence (AGI) presents transformative benefits and risks.The approach focuses on addressing two areas of risk: misuse and misalignment.To prevent misuse, the strategy includes identifying dangerous capabilities and implementing security measures.To address misalignment, model-level mitigations and system-level security measures are proposed.