Trust Center

AI Governance

AI accelerates scoring, but it does not issue credentials on its own. This document explains how AI is used inside Competency Labs and where human review is required.

Where AI is used

Pre-scoringAI produces an initial competency and entrustability score against the published rubric.
Feedback generationAI produces the development-mode feedback shown to learners.
Integrity signal reviewAI helps surface anomalous session and timing signals for human reviewers.

Where AI is not used

Passport issuanceNo Passport is issued from an AI score alone. Every verified Passport requires human reviewer approval.
Identity decisionsIdentity confirmation and revocation decisions are made by trained reviewers, not models.
Selling your dataYour responses are not used to train third-party foundation models. Provider agreements prohibit training on our inference inputs.

Model boundaries

Models operate against a fixed, versioned rubric and structured scoring schema. Prompts, rubric versions, and model versions used for a given assessment are recorded so a score can be reconstructed and audited. Rubric changes are versioned and do not retroactively alter previously issued Passports.

Human review

Verified evidence is reviewed by a qualified human reviewer against the same rubric before it is added to a Passport. Reviewers can accept, adjust, or reject the AI pre-score and are required to record the basis for material adjustments.

AI-only development results are labeled as such throughout the product and are never published as verified Passport evidence.

Maintained by Competency Labs. Contact support@competencylabs.com for corrections or questions.