Evoked · AI-governance course

Is your AI governed, or just instructed?

Five questions, honest answers. It does not assume you fail - plenty of teams find they are further along than they thought, and it says so. It runs entirely in your browser: no account, no sign-up, nothing sent anywhere.

Served under evoked.dev's Sovereign Article Protocol: no tracking, no third-party requests, no beacons. Your answers never leave this page - the score is computed in your own browser. Open your network tab and check.

Pick the honest answer, not the aspirational one. Written for anyone deploying AI agents; if a term like "allow-list" or "the model" is unfamiliar, the free course defines each one.

1Can the agent take an action you never explicitly permitted? No - it can only do what it was allowed Yes - or I am not sure

2Is your limit enforced outside the model, or only written in the prompt? Enforced outside the model (code, allow-list, approval) Written in the prompt / system message

3If the model and your rule disagree, does the rule win by design, or by luck? By design - the rule is enforced, not requested By luck - it usually behaves, but nothing guarantees it

4Can the person the agent acts on see a decision it made about them? Yes - there is a record they can see No - the reasoning stays inside the system

5Can that person contest a decision in a way that reaches a human? Yes - there is a real path to a person No - or only an automated loop

Answer all five to see your standing.