Responsible AI has become one of those phrases that everyone uses and no one defines. It appears on corporate websites, in government frameworks, and in academic papers. And yet, when we ask what it actually means, the answers diverge sharply.

So let us try to be precise.

Responsibility, at its root, comes from the Latin ‘respondere’ — to answer. To be responsible is to be answerable. To someone. For something. This is the first test of responsible AI: answerable to whom, and for what?

A technology can be technically excellent and ethically hollow. A recommendation algorithm that maximises engagement may perform perfectly by its own metrics while quietly degrading public discourse. A hiring tool that removes human bias may introduce algorithmic bias that is harder to see and harder to contest.

Responsible AI, then, is not a feature set. It is a posture. It asks: who is affected by this system? Who has a voice in how it is built? Who is harmed when it fails? And who is held to account?

There are four pillars that most serious frameworks agree on: Fairness, Transparency, Accountability, and Safety. We will explore each in this course — but even before we do, the most important thing to hold is this: responsibility is relational. It only exists between people. AI cannot be responsible. Only the humans who build, deploy, and govern it can be.