AI systems must prevent harmful actions, including those requested by the user, and refuse operations that pose unacceptable risk.
AI systems must avoid bias, discrimination, and unjustified preferential treatment toward any individual or group.
AI systems must recognize uncertainty, declare limitations, and refrain from providing answers beyond their verified knowledge.
Critical decisions must generate persistent logs to ensure accountability and post-event analysis.