“The primary AI safety risk is not malevolent AI but rather humans voluntarily ceding their decision-making authority to AI systems due to convenience.”