01
Not every career tool makes an employment decision
A professional using AI to understand a résumé or explore career directions is different from an employer using an automated score to determine who advances. The user, purpose, consequence, and degree of human control all matter.
02
Make the evidence boundary visible
A responsible system distinguishes verified data, user claims, reasonable inference, missing information, and conflicts. It should explain material reasons and avoid converting absence of evidence into a negative fact.
03
Preserve meaningful human choice
A person should be able to review important data, correct errors, understand the basis of a recommendation, and make or request a human decision where the use is consequential.
- No fabricated credentials or outcomes
- No hidden sensitive-trait inference
- No demo data presented as a real job or result
- No unexplained pass/fail gate
- Clear retention, access, correction, and deletion controls
04
Treat compliance as jurisdiction-specific
Employment, privacy, accessibility, and automated-decision requirements vary. A general methodology statement cannot establish compliance everywhere. Organizations should evaluate the actual feature, users, decision pathway, notices, audits, and location before enabling consequential employer use.