What does 'defensible' mean in the context of AI governance?

It means the organization can produce, on demand, the evidence required to defend any AI-driven decision — the prompt, the retrieved context, the model, the output, the operator, the timestamp, and the policy under which it ran. Defensibility is the practical test of governance.