Skip to main content

What is AI governance?

The AI Glossary

In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.

AI governance is the set of decisions, responsibilities and controls through which an organization directs its use of AI. It establishes who may approve a use, what evidence is needed, who monitors it and what happens when it does not work as intended.

Ethical principles describe what matters. Governance makes those principles operational. An organization may value privacy, for example, but still needs to decide which information a particular tool may receive, who can authorize access and how that decision is checked.

A practical starting point is an inventory of AI-enabled services, including experiments and externally purchased tools. Each entry should have a defined purpose, an accountable owner and a description of the people affected. The inventory is useful only if it changes when the service changes.

Review should be proportionate to consequences. A tool that helps draft an internal agenda does not need the same process as one influencing access to treatment. But calling a tool administrative should not exempt it from scrutiny: billing, scheduling and coverage decisions can have serious effects.

Governance also covers change. A new model version, prompt, data source or permission can alter behavior even when the product name stays the same. Teams need criteria for deciding which changes require renewed evaluation and who has authority to pause or withdraw the service.

This is not solely a committee exercise. People doing the work need a route to report problems, obtain decisions and recover from mistakes. A process that produces approval paperwork but leaves unresolved cases without an owner is incomplete.

For a chronic-disease outreach assistant, governance might specify that it can read designated registry records, draft language-concordant check-in messages and route clinical concerns to a nurse. The service owner would monitor outreach completion, errors and patient feedback, and have authority to restrict actions when problems emerge. Good governance makes experimentation possible within clear boundaries. It should help teams make better decisions throughout use, rather than serve as a one-time ceremony before deployment.

Sources: NIST — AI RMF Core: Govern, Map, Measure and Manage; WHO — Ethics and Governance of Artificial Intelligence for Health (2021)

To learn more, take a look at the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.


AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.

For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.

Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy

Lead with AI in Health Care

The AI Pragmatist is a monthly newsletter from ALA faculty, with new explainers & essays from the Advanced Learning Academy