Safeguard Documentation Center
Safeguard Academy

Exams & Certifications

How Academy certification exams work — question-set and AI-interview modes, the grading pipeline (auto, AI with human review), cooldowns, and certification requirements.

Exams & certifications

Certifications and requirements

A certification defines what must be satisfied before a certificate is issued — typically completing its course and passing its exam. When every requirement is met, the certificate is issued automatically and you receive an email with your public certificate link.

Certifications carry a validity period (the Safeguard Certified Practitioner is valid for 24 months from issue). Expiry dates appear on your dashboard, your calendar, and the public certificate page.

Exam modes

The exam engine supports multiple modes; each certification states which it uses.

Question set (timed)

The standard mode, used by the Safeguard Certified Practitioner exam:

  • Questions are drawn randomly from a larger pool per attempt — no two attempts see the same set.
  • The timer is server-authoritative. For the Practitioner exam: 30 questions, 45 minutes, 70% to pass (21 of 30).
  • Question types include multiple choice, multi-select, and true/false. Multi-select is graded exact-match — no partial credit.
  • Navigate freely between questions and flag questions to revisit; answers are saved as you go.
  • If you disconnect, resume the attempt — the clock keeps running. If time expires mid-attempt, the answers you recorded are submitted automatically.

Subjective questions

Exams can include written-answer questions. These are graded by AI against a rubric, with the grading rationale stored, and escalate to human review when confidence is low or on appeal. Your attempt shows a grading status until every answer has a final score; you're emailed when the result lands.

AI-conducted interview

For certifications that use it, the exam is an adaptive, oral-style interview conducted in chat by the certification's AI tutor — questions adapt to your answers within a fixed turn and time budget. The full transcript is recorded, the AI produces rubric scores with rationale, and human moderation can review before the result is finalized.

Results, cooldowns, and retakes

  • Results show your score and a per-domain breakdown (for example: fundamentals, scanning, compliance, platform) — never per-question answer keys.
  • A failed attempt starts a cooldown (24 hours for the Practitioner exam) before you can retake. The cooldown end shows on your calendar.
  • Passing an exam that completes a certification's requirements issues the certificate immediately.

Integrity

  • Answer keys, rubrics, and model answers never leave the server.
  • AI grading is explainable — the rationale is stored — and human-overridable, with overrides audit-logged.
  • Exam draws are randomized per attempt, and submission is idempotent (a double-submit can't double-grade).

On this page