SkillCort Trust Center
How SkillCort uses and governs AI
AI helps prepare and support an evaluation. People keep ownership of the outcome, and how AI was used is recorded.
Where SkillCort uses AI
Depending on how an organization configures an assessment, AI may assist with:
- drafting assessment content,
- rubric-supported evaluation that a human confirms,
- interview and simulation assistance,
- transcription or analysis where applicable,
- integrity-signal analysis,
- staff-triggered review of proctoring webcam images that flags frames for a human.
Not every feature always uses AI; it depends on the assessment configuration.
Where AI does not have final ownership
- AI does not own final scores.
- AI does not own workflow decisions.
- Integrity signals are reviewed by a human before any decision.
- AI output remains advisory and separately tracked.
Provenance and versioning
- The model identity is recorded where AI is used.
- Relevant configuration and prompt context are versioned.
- AI generations can be traced to their configured context.
Internal prompt content and proprietary configuration are not exposed publicly.
Human oversight
- Authorized evaluators own outcome changes.
- Changes are recorded in a versioned history.
- Review and incident controls exist.
AI pause and incident handling
Critical AI functions can be paused, and incidents can be routed for review. A pause stops the AI provider call and never scores or rejects a candidate. These public pages do not disclose operational security details.
Third-party AI providers, by name
Where AI features run on external providers, we say which ones:
- OpenAI— candidate speech in AI-guided sections (speech understanding and the interviewer's voice), transcription, and the staff-triggered proctoring review of session media.
- HeyGen— the interviewer's on-screen presenter, both the pre-rendered question clips and live sessions. It receives the interviewer's lines to animate — never the candidate's audio or camera. The question clips are rendered from authored question text alone and carry no candidate data at all.
- Anthropic — staff-side drafting and evaluation assistance (question drafting, rubric suggestions, review summaries).
Verdicts, scores, recordings and candidate records stay within SkillCort; providers receive what the feature needs and nothing else. The formal subprocessor list ships with the legal documents currently in preparation.
Candidate disclosure
Candidates are informed when AI is used in an assessment.
The notice shown before an AI-guided assessment names the providers above and states what each receives.
Candidate-facing participation acceptance and decline controls are not currently active.
Planned candidate participation policy
For assessments that require AI processing, the planned behaviour is that declining would stop assessment execution — no score, failure result or automated adverse decision would be created, and no AI-free version would be guaranteed. Production enforcement of this remains a future implementation and is not current behaviour.
Known limitations
The full list of verified controls and documented limitations is in the public governance report.
Read the public governance reportMore in the Trust Center
- Human Decision ControlPeople own assessment outcomes; AI outputs stay advisory.
- Candidate RightsWhat the candidate experience is today, described honestly.
- Security & PrivacyHow SkillCort approaches privacy-by-design, access, retention and backups.
- AccessibilityThe accessibility target and current state, without overstating conformance.
- Governance ReportA sanitized, versioned report of verified controls and documented limitations.
Public Trust Statement v1.8.0 · source governance manifest v1.6.0 · last verified 5 August 2026 · Closed with documented limitations.