WHO Guidance on AI in Health

The World Health Organization’s guidance on ethics and governance of AI for health sets out principles for designing, developing, and using AI in ways that protect people and public interests. Its framework calls for human autonomy, safety and well-being, transparency, responsibility and accountability, inclusion and equity, and responsiveness and sustainability. These principles apply across the lifecycle, from deciding whether AI is appropriate to monitoring how it works in practice. WHO emphasizes that people should remain in control of healthcare systems and medical decisions, and that privacy and confidentiality should be protected. AI should be developed for well-defined use cases, with quality controls and meaningful information about design and limitations. Organizations should establish mechanisms for questioning decisions and redress when people are harmed. Inclusion requires attention to who benefits, who is left out, and how access is distributed. The guidance is normative and does not certify a product or replace national regulation. Health systems can use it to build governance: define accountability, engage patients and health workers, assess bias and safety, protect data, and monitor real-world use. WHO’s later guidance on large multimodal models applies similar principles to generative AI, urging careful evaluation and human oversight. Governance should be adapted to local law and health-system capacity. Consider procurement, workforce effects, and how patients can opt out or ask questions. Assign responsibility for responding when a system behaves unexpectedly.