ZymbosAdvisory Back to site
Illustrative sample only. The system, figures and findings below are fictional and do not describe a real client or assessment.

Model Card

A one-page, shareable summary of an AI system and how it was assessed
PREPARED BY
Zymbos Advisory
Independent assessment
System
TalentSift CV-screening assistant
Reference
ZA-SAMPLE-0001
Version assessed
v2.3
Assessment date
June 2026
Provider / owner
Sample Co Ltd, People Operations
Platform
oxethica
Risk score
3 / 4
High-risk
Confidence score
78 / 100
Evidence gaps remain

Classification

High-risk under the EU AI Act, Annex III (employment and worker management), because the system materially influences shortlisting decisions about people. Also mapped to: UK (ICO guidance and the principles-based regime), US (NIST AI Risk Management Framework and New York City Local Law 144 bias-audit duty), and the equivalent expectations in the UAE, Canada and Australia.

Intended purpose

In scope
  • Rank and shortlist applicants against a structured role profile.
  • Surface relevant experience for a human recruiter to review.
  • Assist a human hiring decision, never make it alone.
Out of scope
  • Automated rejection without human review.
  • Inferring protected characteristics or emotions.
  • Use outside the assessed role families.

System overview

AttributeDetail (illustrative)
Model typeGradient-boosted ranking model over structured CV features, with a text encoder for free text.
Training data180,000 historic applications (2019 to 2025), de-identified, with documented lineage and lawful basis.
Human oversightA recruiter reviews every shortlist; overrides are logged; no automatic rejection.
TransparencyCandidates are told AI assists screening; an explanation is available on request.

Key metrics

MetricResultThresholdStatus
Accuracy (top-20 precision)0.81at least 0.75Pass
Demographic parity ratio (gender)0.86at least 0.80Pass
Equal-opportunity gap (ethnicity)0.110.10 or lessWatch
Data-governance record completeness72%100%Gap

Top risks and mitigations

RiskMitigationOwner
Historic hiring bias amplified by the training dataQuarterly fairness testing, reweighting, protected-group monitoringPeople Ops
Recruiters over-relying on the rankingMandatory review step, override prompts, periodic audit of overridesTalent lead
Incomplete data-governance evidenceClose documentation gaps to lift Confidence above 90Data owner

Limitations

Assessed for the stated role families and regions only. Performance on roles or populations outside the training distribution is not assured. Re-assessment is required after any material change to the model.

Status and next review

Conditional  Cleared for continued use provided the two evidence gaps above are closed within 60 days. Next scheduled review: December 2026, or on any material change, whichever is sooner.

Illustrative sample produced by Zymbos Advisory to show the structure of a Model Card. All names, figures and findings are fictional. Final outputs are produced on the oxethica platform and provided as Zymbos Advisory deliverables. Assessments support your conformity declaration and are not legal advice. © 2026 Albany Limited, trading as Zymbos Advisory.