Ethical AI systems must learn not to discriminate or manipulate anyone based on race, color, religion, gender, age, national origin, marital status, social status, genetics, or medical information. Must understand its own misuse or misconduct limits and recertification mechanisms.