Anthropic AI safety leader Mrinank Sharma resigns, warning that “the world is in peril,” citing global crises, pressure on values, and the pace of AI.