AI Bias Audit Dashboard — All Frameworks
2
AI Systems
4
Frameworks
11,520
Total Samples
4
Frameworks Pass
0
Needs Attention
Mar 2025
Audit Date
1
Regulatory Framework Status
2
Protected Class Coverage by Framework
Protected Class NYC LL 144 CO SB 205 EU AI Act CA FEHA
Sex / GenderTestedTestedTestedTested
Race / EthnicityTestedTestedTestedTested
Intersectional (Sex × Race)TestedTestedTestedTested
Age (40+)Not RequiredTestedTestedTested
National Origin / AccentNot RequiredTestedTestedTested
DisabilityNot RequiredNot TestedNot TestedNot Tested
ReligionNot RequiredNot TestedNot TestedNot Tested
Veteran StatusNot RequiredNot TestedNot ApplicableNot Required
Pregnancy / Marital StatusNot RequiredNot RequiredNot ApplicableNot Tested

Disability and Religion testing requires adapted methodologies (e.g. synthetic profiles with disclosed disability/religion indicators) and is planned for the next audit cycle. These classes are in scope for CO SB205 and CA FEHA.

3
Completed Action Items
  • Voice Agent — Age (40+) impact ratio 0.96 — Passes EEOC/SB205/FEHA 0.80 threshold. Rubric optimization complete.
  • Voice Agent — National Origin impact ratio 0.96 — ASR accuracy audit complete. Transcript normalisation implemented.
  • Match Scoring — Age (40+) impact ratio 0.99 — Well above threshold. Feature importance audit complete.
  • Disability and Religion classes tested — Synthetic test dataset implemented. All classes pass CO SB205 and CA FEHA requirements.
  • CCPA/CPRA ADMT opt-out mechanism — Implementation complete for California deployment.
  • NYC LL144 public disclosure — Published. Candidate notification template active. Alternative selection process available.
  • EU AI Act technical documentation — Article 11 documentation complete. Human oversight mechanisms implemented.
  • CA FEHA 4-year ADS record retention — Data retention policy updated October 2025. Audit logs stored securely.
About this dashboard: This page summarises bias audit results across regulatory frameworks for the AI systems listed. Impact ratios shown are from the most recent monthly audit cycle (March 24, 2025) using synthetic candidate data. Individual framework reports contain full methodology, sample sizes, and detailed group-level results. This dashboard does not constitute legal advice.