Rules That AI Should Follow
What rules should AI follow? ==AI should be useful, safe, fair, transparent, accountable, and respectful of human rights.== [:cite[1]{ln=1}] 1. Protect people’s rights. AI should respect privacy, equality, dignity, ...
What rules should AI follow? ==AI should be useful, safe, fair, transparent, accountable, and respectful of human rights.== [:cite[1]{ln=1}] 1. Protect people’s rights. AI should respect privacy, equality, dignity, freedom, and the rule of law. [:cite[2]{ln=1}] [:cite[3]{ln=3}] 2. Avoid discrimination. Systems should be tested for biased or discriminatory outcomes, especially when decisions affect access to benefits, jobs, health care, education, or justice. [:cite[1]{ln=1}] [:cite[4]{ln=1}] 3. Keep humans responsible. AI should support—not automatically replace—human judgment in high impact decisions, and people should be able to review or appeal harmful decisions. [:cite[5]{ln=3}] [:cite[6]{ln=3}] 4. Be transparent. People should be told when AI is being used, what role it plays, and how important decisions can be questioned or explained. [:cite[4]{ln=1}] [:cite[7]{ln=1}] 5. Protect personal data. AI should collect and use data lawfully, securely, and only for legitimate purposes, with appropriate privacy safeguards. [:cite[8]{ln=1}] [:cite[1]{ln=1}] 6. Be safe and reliable. Systems should be tested for performance, robustness, cybersecurity, and security risks before and during deployment. [:cite[4]{ln=1}] [:cite[9]{ln=2}] 7. Remain accountable. Organizations using AI should set clear performance standards, keep audit trails, monitor results, investigate failures, and correct problems. [:cite[10]{ln=6}] [:cite[11]{ln=4}] 8. Be monitored over time. AI systems can deteriorate as data, circumstances, and security threats change, so they require continuous monitoring, updates, and active maintenance. [:cite[12]{ln=3}] [:cite[12]{ln=5}] [:cite[13]{ln=4}] 9. Respect human control. AI should not be used for clearly unacceptable purposes such as manipulation, discriminatory systems, or biometric identification without consent. [:cite[14]{ln=1}] [:cite[14]{ln=3}] 10. Follow adaptable rules. AI governance should use existing laws where possible, identify gaps, involve affected communities, and update rules as the technology changes. [:cite[15]{ln=3}] [:cite[16]{ln=7}] [:cite[17]{ln=4}] ==In short: AI should help people without hiding its role, violating their rights, making unreviewable decisions, or operating without oversight.== [:cite[7]{ln=1}] [:cite[18]{ln=4}]