What is responsible AI?

Responsible AI means building and using AI in ways that are fair, safe, transparent, respectful of privacy and accountable. Many companies and governments publish responsible AI principles. In practice it means specific controls. These include checking results for bias, keeping people in charge of important decisions, recording what the AI did, protecting data, and telling people when they are dealing with AI.

Why it matters for a business

Principles on a web page change nothing on their own. What protects a business and its customers is specific work underneath: deciding what happens to a wrong answer before anyone acts on it, reviewing outputs across the groups a decision affects, keeping records that let a decision be explained, and knowing where the data travels. Done project by project, that work is responsible AI whether or not anyone calls it that.

Go deeper

Human in the loop and guardrails: how much checking AI needs

Related terms

• AI governance
• AI bias
• Explainable AI
• Human in the loop
• AI guardrails

All terms, A to Z