Product
GovSafeAI Toolkit
Templates, playbooks and guided journeys for responsible AI in the APS.
The practical instruments for standing up responsible AI in a government agency, including the parts other toolkits leave out: the forbidden questions, the coalition-building, and the political survival of an AI programme.
Capabilities
Here is what GovSafeAI Toolkit does.
You start from a template rather than a blank page
AI Impact Assessments, registers, risk tiers and review thresholds as working documents you adapt, rather than principles you have to translate into instruments yourself.
We sequence the work so you know what comes first
Sequenced paths through the work, so an agency starting from nothing knows what to do first and what can safely wait.
We cover the parts other toolkits leave out
Forbidden questions, coalition-building and political survival. The organisational realities that determine whether an AI programme lives, and that no framework document mentions.
Everything maps to the obligations you answer to
Mapped to the whole-of-government AI Policy, the APS AI Plan, the PSPF and the Privacy Act, so the outputs answer to the frameworks you are measured against.
Status
The GovSafeAI Toolkit is releasing soon. If you are standing up AI governance now and would like early access, or would like to shape what goes into it, we would like to hear from you.
Why we are building it
Every governance engagement we run produces the same artefacts: an impact assessment process, a register, risk tiers, review thresholds, an acceptable-use position, a briefing that gets an executive over the line. We have drafted these enough times to know which decisions recur and which are specific to an agency.
Publishing them means an agency without the budget for a governance engagement is not starting from a blank page. That is deliberate. Most of the value we add is in judgement and sequencing, not in withholding a template.
The part that is usually missing
Governance frameworks describe what good looks like. They rarely describe how to get there inside an organisation that has competing priorities, a sceptical executive, a nervous legal team and a change of leadership pending.
That gap is where AI programmes actually die. So the toolkit covers the questions people are reluctant to write down: which questions you must not ask in front of which audience, whose support you need before a paper goes up, how to sequence a proposal so it survives contact with a budget cycle, and how to keep a programme alive across a change of minister or Secretary.
None of that is in a policy document. All of it determines whether the policy document ever gets implemented.
Next step
We can show you GovSafeAI Toolkit running against your own material.
A demonstration using your documents and your terminology tells you more in an hour than a scripted walkthrough does.