AI Governance Practice
Practice the judgment behind responsible AI decisions.
A visual, scenario-based study aid for people who want to get better at turning
AI governance principles into practical action.
What is this?
AI governance is not just knowing the vocabulary. It is knowing what to notice,
who is responsible, what could go wrong, and what to do next.
AI Governance Practice gives you realistic situations and asks you to make
the call. You choose an answer, see the reasoning, and build a repeatable way of
thinking that carries into real governance work.
| | |
|---|
| 350 | 4 | 0 |
| original questions | governance passes | accounts required to start |
The four-pass method
Every scenario can be read through the same four passes. This keeps you from
jumping to a familiar framework before understanding the situation in front of
you.
| Pass | The question it asks |
|---|
| 01 · Facts | What is actually happening? |
| 02 · Obligations | What is required, and by whom? |
| 03 · Risks | What could harm people or the organization? |
| 04 · Action | What is the narrowest defensible next step? |
| Step | What you do |
|---|
| 01 · Notice | Read the situation without jumping to a framework. |
| 02 · Reason | Separate obligations from assumptions. |
| 03 · Choose | Pick the narrowest defensible response. |
| 04 · Learn | Review the rationale and carry the lesson forward. |
The app is intentionally calm and readable: one scenario, four answer choices,
clear feedback, and enough explanation to understand why an answer is
defensible. Progress begins in the browser, so you can start immediately
without signing up.
What you can explore
The current track contains 350 original questions across four areas:
| Track | What it gives you |
|---|
| Foundations of AI Governance | Shared vocabulary, roles, accountability, and governance structure. |
| Laws, Standards, and Frameworks | A way to distinguish obligations, standards, and voluntary guidance. |
| Governing AI Development | Design choices, evaluation, controls, and risk reduction. |
| Governing AI Deployment and Use | Oversight, transparency, incidents, monitoring, and impact. |
Of those, 62 hang off 15 multi-question fact patterns —
one organization, one system, one set of competing pressures, interrogated from
several angles. The remaining 288 are standalone items written in the same
situational style. "Scenario-based" describes how the questions read; the
fact-pattern count is the smaller, separate number.
The content touches responsible AI, risk identification, accountability,
transparency, explainability, human oversight, fairness, monitoring, incidents,
lifecycle controls, and practical governance decisions. It is built for
judgment — not rote memorization.
How it works as a product
Focused study sessions, domain practice, a review queue, progress tracking, and
a timed practice sitting. The motivational layer stays deliberately light: a
daily goal, a streak, and the satisfaction of getting better at the hard
questions.
Small loop. Better judgment. Repeat.
The interface is mobile-first, works without an account, and keeps progress
locally on the device. Optional sign-in adds synchronization across devices.
Run it locally
npm install
npm run dev
Then open http://localhost:3000. No environment variables are required for
the core study experience.
npm run build # server build — local preview only
npm run build:pages # static export for GitHub Pages
npm run build:mobile # static export for the Capacitor app
npm run check # lint, typecheck, tests, and content checks
npm test # unit tests
Practice exam mode
Exam mode is an independent timed simulation over the same original
questions. It samples from the bank, so no two sittings are identical — it
is not a fixed form.
- Multi-select grades all-or-nothing; unanswered questions count as incorrect.
- Options are dealt fresh letters each sitting, and correctness is stored against
option identity, so no answer can be memorised as a letter.
- The clock scales with length at a constant pace: 25 questions in 45 minutes,
50 in 90, 100 in 180.
[!NOTE]
That pace is this project's own choice. It has not been verified against
any official published figure and is not presented as one. Exam mode does not
reproduce any certification exam and does not predict a result.
Content quality
Question changes are governed by
docs/CONTENT-QUALITY-CONTRACT.md and
enforced by npm run check:content, which fails on missing metadata, broken
answer keys, degenerate multi-select, unlocatable sources, missing or stub
distractor notes, skewed answer positions, length bias, similarity, and stale
coverage documentation.
npm run check:content # content invariants
npm run release:gate # the above + lint, types, tests, both builds
Deployment
The app is a static export — no server, no API routes, no middleware — so
GitHub Pages hosts all of it. Supabase is called straight from the browser, and
its Edge Functions are hosted by Supabase.
Pushing to main runs
.github/workflows/deploy-pages.yml,
which builds and publishes.
- Set Settings → Pages → Source to GitHub Actions once.
- Optional
NEXT_PUBLIC_* values go in Settings → Secrets and variables →
Actions → Variables — variables, not secrets. Every NEXT_PUBLIC_* value
is inlined into the browser bundle at build time, so none of them may hold a
secret, and changing one needs a new workflow run to take effect.
Full notes, including the base-path rules a project site imposes, are in
docs/mobile-release.md.
Original content
All questions, scenarios, rationales, and takeaways in this project are
independently written educational material. They are not copied from
certification examinations, proprietary question banks, or official training
resources.
The project discusses public frameworks, standards, and regulations — including
the NIST AI Risk Management Framework, the EU AI Act, and ISO/IEC 42001 — only
as subject matter for learning. Mentioning them does not imply affiliation,
endorsement, certification, or formal compliance.
[!IMPORTANT]
Study-aid disclaimer. This is a free supplemental study aid for
professional development. It is not an official certification resource,
course, training program, or substitute for primary preparation materials. It
is not affiliated with or endorsed by any certification organization. It does
not contain actual certification examination questions and does not guarantee
examination success.
Contributing
Found a bug? Have an accessibility improvement? Want to suggest an original
governance scenario? Open an issue and describe what you noticed.
[!WARNING]
Please do not submit recalled, copied, confidential, or proprietary
examination material.
Project status
The live product is available at
thankcheeses.github.io/ai-governance-practice.
The repository is being actively shaped as a public portfolio project, with new
content and visual refinements added as the learning experience evolves.
Built to make governance judgment feel practiceable.
Start a scenario →
AI Governance Practice is an independent educational project.
Framework and organization names belong to their respective owners.
