tl;dr
Accountable AI refers to AI systems built so their decisions can be explained, audited, and traced back to a specific action and reason, rather than operating as a black box that produces an output without a visible reasoning trail.
For finance, where every number needs to be defensible to an auditor or a regulator, this isn't optional. An AI system that codes a transaction or flags an exception needs to show its work: what it saw, what rule or pattern it applied, and who's responsible for reviewing it.
Back to the glossary