Written for the people who have to answer for an AI system: executives, risk and audit committees, assurance specialists and the teams building the systems. Guides, working checklists and a free diagnostic — no registration required to read or download.
Most organisations have AI principles and a committee. Far fewer can produce the dated evidence that shows a specific AI system was assessed, reviewed by someone independent, and approved by a named person. That difference is AI assurance.
Benchmark scores describe a model. Assurance needs evidence about your application of it — on your data, against your thresholds, with a result someone can re-run. This guide covers how to design evaluations that stand up as evidence.
What assurance actually requires beyond principles: contextual risk classification, matched control obligations, dated evidence, independent review and a recorded readiness decision.
Read: What is AI assurance?Five evaluation disciplines — accuracy and grounding, safety and guardrail resilience, fairness, human oversight agreement and production drift — and how to record results so they count as evidence.
Read: AI evaluationsHow the Voluntary AI Safety Standard guardrails, the Privacy Act and Australian Privacy Principles, ISO/IEC 42001 and the NIST AI RMF fit together across one body of assurance evidence.
Read: The Australian contextBoth downloads are branded PDFs generated in your browser. Nothing is submitted and no email address is required.
Ten questions to ask before an AI system reaches production
A short working checklist for executives, risk owners and boards. Each item is a question you should be able to answer from a record rather than from recollection.
A structure for evaluations that stand up as assurance evidence
A template for teams designing evaluations of an AI system. Complete one protocol per evaluation, before running it, and keep the completed protocol with the result.
Answer a short set of questions about one AI system and receive an indicative risk tier, a governance defensibility read, the number of control obligations that system carries, your top blindspots and a recommended next step — as an on-screen review and a PDF report.
