Glossary term
Responsible AI
Subject: Technology, Cyber and Data, AI Governance and Risk
Also known as: Trustworthy AI, AI Ethics
Responsible AI is the practice of designing, building and using AI according to principles such as fairness, safety, transparency, privacy and accountability, and turning those principles into working controls.
Explanation
Most organisations begin with a statement of principles, often drawn from sources such as the OECD AI Principles or from their own ethics board. The harder part is converting words into practice: impact assessments, bias testing, documentation, human oversight, review boards and training for builders. The terms overlap. Trustworthy AI, used by the EU and NIST, stresses that systems earn trust through properties like reliability and safety. AI ethics tends to describe the underlying moral questions, such as what AI should never be used for. Responsible AI is the working discipline that links both to daily processes. Critics warn of ethics-washing, where a published code exists but no decision is ever changed by it. The test of a programme is whether any project was stopped or changed on ethical grounds, and who had the authority to do it.
Example
A retailer publishes five AI principles. To make them real, it requires every new AI project to pass a review panel covering data sources, bias tests, privacy and a named owner. The panel rejects a plan to infer customers' health conditions from their purchases, judging it intrusive, and the project team redesigns the tool to use only information that customers opt in to share.
Not to be confused with
- AI Governance
- Responsible AI is the set of values and practices; AI governance is the structure of roles, approvals and oversight that holds them in place.