AI Fairness Transparency Report

ZT Nation's commitment to unbiased AI-powered contractor matching

Our Commitment

ZT Nation uses AI to match customers with qualified contractors. We are committed to ensuring our AI systems operate fairly and without discrimination. We monitor all 9 federal protected classes, conduct automated bias audits every 30 days, and maintain impact ratio thresholds that exceed EEOC requirements by 12.5%. Our five-layer self-healing architecture automatically detects and corrects bias drift before it can affect outcomes.

This page is published in compliance with NYC Local Law 144, Colorado SB 24-205, and EEOC guidance on AI in employment decisions.

Current AI Fairness Status

Compliant

Last Audit

Pending first audit cycle

Decisions Analyzed

0

Audit Frequency

Every 30 days

Protected Classes Monitored

ZT Nation monitors all federally protected classes in our AI scoring systems. Our AI does not collect, store, or use any demographic data. Instead, we use geographic proxy analysis to detect potential disparate impact.

Race

Title VII, Section 1981

Color

Title VII

Sex (incl. sexual orientation, gender identity)

Title VII, Bostock v. Clayton County (2020)

Religion

Title VII

National Origin

Title VII

Age (40+)

ADEA

Disability

ADA

Genetic Information

GINA

Pregnancy

PDA

Impact Ratio Thresholds

We use the EEOC four-fifths rule as our baseline, with ZT Nation's premium standard set 12.5% higher. Impact ratios measure whether all groups receive proportionally equal treatment in our AI matching system.

ZT Premium Standard90%+

Compliant. Documented in monthly report.

Early Warning85-89%

Investigation within 7 days. Root cause identified.

Elevated Concern80-84%

Root cause analysis within 48 hours. Executive notification.

Critical / EEOC ThresholdBelow 80%

Immediate system pause. AI scoring halted. Manual review required.

The EEOC four-fifths rule sets the legal threshold at 80%. ZT Nation's premium standard of 90% provides an additional 10-percentage-point buffer above the legal minimum.

Five-Layer Self-Healing Architecture

ZT Nation's AI systems include automated safeguards that detect and correct bias drift without human intervention. This architecture runs continuously and is modeled after methodologies from ORCAA, Holistic AI, BABL AI, and Credo AI.

1

Continuous Monitoring

Every scoring decision is logged and analyzed in real-time. Impact ratios computed by geographic proxy groups every 15 minutes.

2

Threshold Correction

When impact ratios drop below 90%, selection thresholds are automatically adjusted per group to restore equal selection rates.

3

Weight Recalibration

At 88% impact ratio, scoring weights are automatically recalibrated. Maximum 3% single correction, 15% cumulative per 30-day period.

4

Circuit Breakers

Below 80%, AI scoring is automatically paused for the affected group. Contractors matched via FIFO queue until resolved.

5

Root Cause Diagnosis

Automated analysis identifies the source of drift and generates corrective recommendations with confidence scoring.

Circuit Breaker Status

Circuit breakers automatically pause AI scoring for any group where the impact ratio falls below 80%. When tripped, contractors are matched via a first-in-first-out queue until the issue is resolved.

No Active Circuit Breakers

All AI scoring systems are operating normally within acceptable thresholds.

Audit Schedule

Continuous

Real-time monitoring of every scoring decision. Automated alerts at 88% threshold.

Monthly (30-Day)

Comprehensive statistical review with intersectional analysis. Automated report generation.

Annual Deep Audit

5-phase independent audit modeled after ORCAA, Holistic AI, BABL AI, and Credo AI methodologies.

Data Practices

No demographic data collected. Our AI does not ask for, store, or use race, sex, age, religion, or any other protected characteristic in scoring decisions.

Proxy variable monitoring. We actively monitor geographic and service-type variables for correlation with protected characteristics using Pearson correlation analysis.

Intersectional analysis. We test combinations of groups (not just individual categories) to detect compounded disparate impact per FTC Rite Aid consent order standards.

Scoring transparency. Contractor scores are based solely on: rating, experience, reliability, proximity, availability, skill match, and trust score. No hidden factors.

Report a Concern

If you believe you have experienced discrimination or unfair treatment through our AI matching system, we want to hear from you. All reports are reviewed within 24 hours. Anonymous reporting is available. Anti-retaliation protections apply to all reporters.

Download Anti-Retaliation Policy (PDF)

Multiple reporting channels: Web form, guided chatbot, SMS (text BIAS to our number), or email. All channels support anonymous submission.

External reporting: EEOC (1-800-669-4000) · FTC · State Attorney General

This transparency report is published pursuant to NYC Local Law 144 (2021), Colorado SB 24-205 (2024), and EEOC Technical Assistance on AI and Title VII (2023).

Methodology references: EEOC four-fifths rule (29 CFR 1607.4D), Bostock v. Clayton County, 590 U.S. 644 (2020), Section 1981 of the Civil Rights Act of 1866.

Last updated: 3/16/2026 | Audit framework modeled after ORCAA, Holistic AI, BABL AI, and Credo AI

We use cookies and similar technologies to provide essential platform functionality and improve your experience. For details, see our Cookie Policy and Privacy Policy.