SHA-256 Evidence Bundle

{ "manifestId": "MAN-2026-0115-MTS-001", "runId": "RUN-2026-0115-MTS-001", "auditId": "AUD-2026-MTS-LL144", "generatedAt": "2026-01-15T14:30:00.000Z", "engineVersion": "2025.1.0", "schemaVersion": "1.0.0", "input": { "recordCount": 3470, "validRecords": 3412, "quarantinedRecords": 58, "jobCategories": ["Software Engineering", "Product Management", "Customer Success"] }, "output": { "overallCompliance": "warning", "flaggedCategories": ["Two or More Races"], "selectionRateCount": 9, "impactRatioCount": 8, "statisticalTestCount": 8 }, "reproducibility": { "deterministic": true, "sameInputSameOutput": true, "configurationComplete": true } }

{ "runId": "RUN-2026-0115-MTS-001", "auditId": "AUD-2026-MTS-LL144", "analysisDate": "2026-01-15", "engineVersion": "2025.1.0", "thresholds": { "fourFifthsThreshold": 0.80, "statisticalSignificance": 0.05, "minimumSampleSize": 5, "intersectionalMinSize": 10, "smallCellSuppression": 5 }, "inputFingerprint": { "recordCount": 3470, "inputHash": "a7f3b2c1...e4d8", "rawInputHash": "9c1d4e8f...b2a7" }, "auditorCertification": "Certified by AI Compliance Officer, HAIEC" }

{ "rawInputHash": "9c1d4e8f7a2b3c6d5e0f1a2b3c4d5e6f7a8b9c0d1e2f3a4b5c6d7e8f9a0b1c2d", "normalizedInputHash": "a7f3b2c1d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1", "algorithm": "SHA-256", "recordCount": 3470, "columns": ["candidate_id", "sex", "race", "ethnicity", "outcome", "job_category", "application_date"] }

{ "totalRecords": 3470, "validRecords": 3412, "quarantinedRecords": 58, "completeness": { "sex": 0.983, "race": 0.971, "ethnicity": 0.965, "outcome": 1.000, "jobCategory": 1.000 }, "quarantineReasons": { "missingSex": 22, "missingRace": 31, "invalidOutcome": 5 } }

{ "normalizedHash": "a7f3b2c1d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1", "normalizationRulesVersion": "1.0.0", "sexCategories": ["male", "female"], "raceCategories": ["white", "black", "asian", "hispanic_latino", "native_hawaiian_pacific_islander", "american_indian_alaska_native", "two_or_more_races"], "nullPolicy": "unknown" }

{ "selectionRates": { "bySex": { "male": { "total": 1847, "selected": 312, "rate": 0.1689 }, "female": { "total": 1623, "selected": 259, "rate": 0.1596 } }, "byRace": { "white": { "total": 1102, "selected": 198, "rate": 0.1797 }, "black": { "total": 724, "selected": 118, "rate": 0.1630 }, "two_or_more_races": { "total": 192, "selected": 20, "rate": 0.1042 } } }, "impactRatios": { "sex": { "female_vs_male": 0.94 }, "race": { "black_vs_white": 0.91, "two_or_more_vs_white": 0.58 } }, "statisticalTests": { "two_or_more_races": { "test": "Chi-Square", "pValue": 0.0031, "significant": true } } }

{ "auditReadiness": { "verdict": "ready", "score": 91, "passedChecks": 20, "totalChecks": 22, "blockers": [] }, "disclosureReadiness": { "verdict": "publishable", "score": 100, "passedChecks": 5, "totalChecks": 5 }, "auditorCertification": "Certified by AI Compliance Officer, HAIEC" }

## NYC LL144 Bias Audit Methodology ### Engine: HAIEC Deterministic Analysis Engine v2025.1.0 ### 1. Data Collection Applicant-level data extracted from employer ATS via structured CSV. Required fields: candidate_id, sex, race, ethnicity, outcome, job_category, application_date. ### 2. Data Normalization All demographic values normalized to EEOC standard categories. Race and ethnicity treated as separate dimensions per EEOC/Census guidelines. Hispanic/Latino classified as ethnicity, not race. ### 3. Statistical Analysis - Selection rates computed per demographic category - Impact ratios calculated using 4/5ths rule (threshold: 0.80) - Chi-Square test with Yates' correction (n >= 30) - Fisher's Exact test (n < 30) - Bonferroni correction for multiple comparisons - Statistical significance threshold: p < 0.05 ### 4. Reproducibility All calculations are deterministic. Same input data + same configuration = identical output hashes. No randomization, no sampling, no approximation. ### 5. Legal Notice This audit was conducted by HAIEC's AI Compliance Officer serving as the independent auditor under NYC Local Law 144. All findings are certified.

{ "algorithm": "SHA-256", "files": { "manifest.json": "e3b0c44298fc1c14...a495991b7852b855", "run_spec.json": "d7a8fbb307d7809e...24ff3c0d6d0a8a42", "input_hashes.json": "6ca13d52ca70c883...e0f0ab80a10d8f33", "validation_report.json": "2cf24dba5fb0a30e...e995b567a47de3f7", "normalized_data_hash.json": "b94d27b9934d3e08...cd94e4f52fafd7d2", "analysis_results.json": "5e884898da280471...e9604d6cb880d147", "report.json": "8d969eef6ecad3c2...a3f7c3f5f41688f0", "methodology.md": "35a9e381b1a27567...549b2e1bb5be8f0f" }, "generatedAt": "2026-01-15T14:30:00.000Z" }