2.1.6. Accountability
💡 First Principle: Accountability means people — not algorithms — remain answerable for how an AI system behaves and is governed. AI can make or influence decisions, but responsibility for those decisions and for the system's oversight stays with humans and organizations.
In practice, accountability means clear ownership: defined governance, people responsible for the system's design and operation, and processes to review and correct it when something goes wrong. It's the principle that prevents "the algorithm decided" from becoming a way to dodge responsibility. Accountability ties the other five principles together — someone has to be answerable for ensuring fairness was checked, safety was tested, privacy was protected, inclusiveness was considered, and transparency was provided.
⚠️ Exam Trap: Transparency and accountability are frequently confused. Transparency is making the system understandable (people can see how/when it's used). Accountability is humans remaining answerable (someone owns the outcomes and governance). A scenario where no one can be held responsible for a harmful automated decision is an accountability gap, even if the system was perfectly transparent.
Reflection Question: A company says "our AI is fully transparent — users can see exactly how it scores them." Does that satisfy accountability? Why or why not?
