Practical Guidance

What Is AI Assurance? A Working Definition, and What It Is Not

AI assurance is evidence that an AI system does what it should, cannot do what it must not, and can be shown to have done either. What that means in practice, and how to tell if you have it.

7 min Read

Updated 2026-09-19

Prompt Injection: Why the Fix Is Permissions, Not a Better Prompt

Prompt injection is not a prompting bug, and it is not patched at the prompt layer. Here is what it actually is, and what genuinely reduces the damage.

8 min Read

Updated 2026-09-10

What an AI Audit Trail Has to Contain to Be Worth Anything

Most AI logging records the answer and throws away everything that produced it. Here is what a trail needs to contain to survive an incident or an audit.

7 min Read

Updated 2026-09-10

Measuring an AI Feature: What to Check Before You Ship It

Twenty prompts that looked right is not a measurement. What an eval set needs to contain, what to measure, and how to stop regressions reaching users.

8 min Read

Updated 2026-09-10