AI trust card
How we use AI, in plain words.
Organised around the four functions of the NIST AI Risk Management Framework (Govern, Map, Measure, Manage) and informed by ISO/IEC 42001. This is a voluntary disclosure, not a certification.
Govern — who is accountable
Platform admins can switch each AI feature on or off, pause it or cap its budget. Admin actions are written to an audit log. Admins cannot use member career tools or read member career data.
Map — where AI is used
AI reads your documents to suggest facts, drafts diagnoses, reviews proof submissions, coaches interview answers, suggests jobs and drafts messages. It never reads your Recovery Space, and voice recordings are discarded after transcription.
Measure — how outputs are checked
Every extracted fact must quote your document word for word and stays pending until you accept it. Confidence levels come from fixed, tested rules, never from the model. A three-reviewer council can only lower confidence, and it shows when the reviewers disagree.
Manage — how people stay in control
You review everything before it counts. Diagnosis versions are kept, not overwritten. Drafted applications are never sent for you. Fit scores, skill levels and proof reviews never predict hiring and never award credentials.
Data boundaries
AI calls run on our servers with your own access rights, so they can only see your records. Sponsors see group totals of five or more people; mentors see only what you share and you can stop sharing at any time.
Known limits
AI can be wrong or incomplete. Job postings come from third parties. Skill ratings are self-reported unless linked to reviewed work. Tell us if something looks wrong and we will fix it.