It will be impossible to implement safeguards for or detect deception from an AI that becomes smarter than humans

Moral principles and responsible-use guidelines shaping how automated systems are built and deployed.

The field of preventing harm from advanced AI systems, spanning alignment, evaluation, and decisions to slow or withhold models whose capabilities outpace their safeguards.

Research topic (Topic).

Practices aimed at minimizing risks in AI systems. Essential for responsible deployment.

Practices ensuring ethical and accountable AI use. Includes fairness and transparency.