Advanced AI capabilities can be leveraged to solve the safety problem: if an AI system can better understand what is right and wrong, it will be less likely to cause harm, making increased capability a tool for reducing risk.