An audit confirms that an AI policy exists on the day someone checks. It says nothing about what the AI did the other 364 days.

Meanwhile, employees use tools nobody approved, vendors switch on models inside software you already run, and your own teams build AI with no record of what went into it. A policy sees none of that, and the missing record is hard to explain after an incident.

This session treats resilience as the test of AI governance. We put two kinds of evidence side by side: what an auditor accepts (a policy, a register, a training record) and what you need when something goes wrong (an inventory of where AI runs, a log of what it did, a guardrail that blocks a violation as it happens, and a tested way to switch a model off).

We'll then cover how to find the AI you did not buy, how to keep an audit trail for the AI you build or use, and how to run a resilience check for AI governance in order to move from documented to demonstrated.