# Harver: HireAIScore fact sheet

URL: https://www.hireaiscore.com/vendors/harver
Fact sheet: https://www.hireaiscore.com/vendors/harver/facts.md
Last reviewed: 2026-11-14
Rubric: v1.0 (https://www.hireaiscore.com/methodology)

> HireAIScore rates Harver 56 out of 100 (grade F, Substantial gaps) under rubric v1.0, last reviewed November 14, 2026. Harver ranks 8 of 82 vendors rated in Assessment AI, against a category average of 46. That score is drawn from 33 evidence items across 7 rubric criteria - 26 with a cited source, 7 recording that nothing was located.

This file is the text of the profile at the URL above, for AI assistants and other software. Scores are an editorial assessment of public evidence against a published rubric. They are not legal advice, a certification, or a statement that a product is or is not compliant.

## Score

- Overall: 56/100, grade F (Substantial gaps)
- Rank: 8 of 82 in Assessment AI (category average 46)

## Criterion scores and evidence

### Article 11 Technical Documentation: 58/100 (weight 20%)

- Public statement (captured 2026-11-14): October 2026 post by Harver's Chief Science Officer describing concurrent and predictive validation studies, a client choice between psychometric and machine-learning models, SHAP explanations, holdout testing and adverse-impact tests. Source: https://harver.com/blog/what-backed-by-science-means-in-hiring/
- Documentation (captured 2026-11-14): Public help article (May 2026) explaining that matching scores are weighted averages of assessment components validated by Harver's People Science team. Source: https://knowledge.harver.com/hc/en-us/articles/19104838623505-Matching-Profile
- Documentation (captured 2026-11-14): Vanta trust centre (78 documents) lists a Responsible AI Development and Deployment Policy, an AI Policy, an AI Impact Assessment Standard and an AI oversight committee charter, all behind Request access; only the form DPA and master SaaS agreement are public. Source: https://trust.harver.com/resources
- Documentation (captured 2026-11-14): Responsible AI page: decision-support framing, claimed alignment with EEOC, OFCCP, NYC LL 144, Illinois and EU AI Act requirements, and a statement that the least-biased candidate model is deployed. Source: https://harver.com/why-harver/responsible-ai-in-hiring/
- Searched, not found (captured 2026-11-14): No Annex IV-style pack, system card, model card or ISO/IEC 42001 statement. Searched: harver.com site search for model card, 42001 and AI Act; harver.com/eu-ai-act and /responsible-ai (404); trust.harver.com/resources; knowledge.harver.com. The two whitepapers previously cited are lead-gated landing pages for a general buyer's guide and a 2024 marketing paper.

### Bias Audit Transparency: 68/100 (weight 18%)

- Audit report (captured 2026-11-14): BABL AI NYC LL 144 audit of the Harver Platform (v1.0 4 November 2025): data April to May 2025, about 280 applicants with known gender; female impact ratio 0.918, lowest intersectional 0.750 on a group of eight. Source: https://harver.com/wp-content/uploads/2025/11/Harver-Platform-2025-Bias-Audit.pdf
- Audit report (captured 2026-11-14): BABL AI NYC LL 144 audit of the pymetrics Soft Skills Platform (v1.0 17 July 2025): data January to December 2024, 283,256 applicants with known gender; lowest race impact ratio 0.914 (Asian). Source: https://harver.com/wp-content/uploads/2025/11/pymetrics-Soft-Skills-Platform-2025-Bias-Audit.pdf
- Public statement (captured 2026-11-14): Harver says it passed an independent NYC LL 144 audit in 2025 and 2026 and keeps an ongoing bias-audit relationship with BABL AI; no 2026 summary is linked or published. Source: https://harver.com/blog/what-backed-by-science-means-in-hiring/
- Public statement (captured 2026-11-14): Independent Stanford, Chapman and Northeastern study (May 2026, FAccT 2026) of 4.2 million pymetrics applications finds clear racial disparities at position level; Harver's Chief Science Officer has publicly rejected its methodology. Source: https://arxiv.org/abs/2605.27371
- Searched, not found (captured 2026-11-14): No 2026 LL 144 summary for either platform. Searched: harver.com/security (links only the two 2025 PDFs), the harver.com media library for audit, bias and BABL files, trust.harver.com/resources (no audit documents), web searches for Harver and pymetrics 2026 bias audits.

### FRIA Support: 35/100 (weight 15%)

- Documentation (captured 2026-11-14): Trust Package (governance, audit readiness, product-specific compliance for Harver, Checkster and pymetrics) delivered on request as a password-protected zip. Source: https://harver.com/blog/harver-trust-package-security-privacy/
- Documentation (captured 2026-11-14): Public EU and rest-of-world master services agreement with a DPA schedule; it contains no AI Act, FRIA or intended-use content. Source: https://harver.com/wp-content/uploads/2025/01/20241212-ONLINE-Harver-Master-Services-Agreement_EU_RoW_v0.7.pdf
- Searched, not found (captured 2026-11-14): No FRIA template, Article 27 deployer guidance or intended-use data sheet. Searched: harver.com/resources, harver.com site search for AI Act, knowledge.harver.com (201 public articles, none on FRIA or the AI Act), trust.harver.com/resources.

### Data Governance Disclosure: 58/100 (weight 15%)

- Documentation (captured 2026-11-14): States Harver's AI never uses demographic data to make decisions. Source: https://harver.com/why-harver/responsible-ai-in-hiring/
- Documentation (captured 2026-11-14): MSA lets Harver use anonymised, aggregated usage data to improve its services and for R&D, and sets candidate data retention to unlimited by default unless the customer sets limits. Source: https://harver.com/wp-content/uploads/2025/01/20241212-ONLINE-Harver-Master-Services-Agreement_EU_RoW_v0.7.pdf
- Documentation (captured 2026-11-14): Candidate data removal rules: Harver recommends 12 months or less; the maximum allowed is 84 months. Source: https://knowledge.harver.com/hc/en-us/articles/14860972042129-Privacy-Settings
- Documentation (captured 2026-11-14): July 2026 norm refresh rebuilt Personality Print norms from 7.4 million candidate records across five regions. Source: https://knowledge.harver.com/hc/en-us/articles/48829687179537-Personality-Print-Norm-Group-Refresh-15-July-2026
- Searched, not found (captured 2026-11-14): No per-model training-data description and no subgroup validity metrics outside the LL 144 selection-rate tables. Searched: the Responsible AI page, the MSA, knowledge.harver.com articles on training, validation and adverse impact, harver.com/privacy-policy (March 2026).

### Human Oversight Design: 58/100 (weight 12%)

- Documentation (captured 2026-11-14): Automation rules change a candidate's status automatically when a condition on time in status, matching bracket or matching score is met; a documented example moves applicants who never finish to Rejected after 30 days. Source: https://knowledge.harver.com/hc/en-us/articles/4412519845777-Automation-Rules
- Documentation (captured 2026-11-14): Candidates can be auto-rejected on hard show-stopper (knockout) questions. Source: https://knowledge.harver.com/hc/en-us/articles/12990768604049-Operation-Insights
- Documentation (captured 2026-11-14): September 2026: result reports can show the relative weight of each indicator (off by default) and candidate strengths and development areas. Source: https://knowledge.harver.com/hc/en-us/articles/51504543697937-Result-Report-Changes-2026-September
- Documentation (captured 2026-11-14): Virtual interview answers are rated by people, not by AI. Source: https://knowledge.harver.com/hc/en-us/articles/41004109799441-Rating-Virtual-Interviews
- Documentation (captured 2026-11-14): Decision-support, not decision-making, framing. Source: https://harver.com/why-harver/responsible-ai-in-hiring/
- Searched, not found (captured 2026-11-14): No public description of a switch to disable AI scoring as a whole or of confidence indicators on match scores. Searched: knowledge.harver.com for override, threshold, human review and disable; the Responsible AI page.

### Post-Market Monitoring: 52/100 (weight 12%)

- Public statement (captured 2026-11-14): Security page now offers a separate 'Report an AI Issue' route (ai-reporting@harver.com) beside GlobalSecurity@harver.com; it was not on the page in May 2026. Source: https://harver.com/security/
- Documentation (captured 2026-11-14): Advance notice of a norm update telling customers pass rates will shift by 1 to 10% for most, with consultant follow-up. Source: https://knowledge.harver.com/hc/en-us/articles/48829687179537-Personality-Print-Norm-Group-Refresh-15-July-2026
- Documentation (captured 2026-11-14): Public dated release-notes series, January to October 2026. Source: https://knowledge.harver.com/hc/en-us/articles/51505778463377-Harver-Platform-Updates-05-OCT-2026
- Documentation (captured 2026-11-14): Live public status page for the legacy Outmatch Assess and Checkster products; not linked from harver.com. Source: https://status.outmatch.com
- Searched, not found (captured 2026-11-14): No public status page for the main Harver Platform and no published drift-monitoring results. status.harver.com and harver.statuspage.io redirect to an inactive Statuspage visible only to team members; harverstatus.com does not respond. Also searched: knowledge.harver.com release notes, the Responsible AI and security pages.

### Customer Documentation: 64/100 (weight 8%)

- Documentation (captured 2026-11-14): Public Harver Platform knowledge base covering vacancy setup, matching profiles, automation rules, report guides, privacy settings, integrations and release notes. Source: https://knowledge.harver.com/hc/en-us
- Documentation (captured 2026-11-14): Includes a warnings section on configuration changes that alter how the matching score is calculated. Source: https://knowledge.harver.com/hc/en-us/articles/19104838623505-Matching-Profile
- Documentation (captured 2026-11-14): Public accessibility and accommodations guide. Source: https://harver.com/wp-content/uploads/2024/08/accessibility-accommodations.pdf
- Searched, not found (captured 2026-11-14): No formal instructions for use, limitations statement or intended-use boundaries per assessment. Searched: knowledge.harver.com, harver.com/assessments, the Responsible AI page.

## Company facts

- Website: https://harver.com
- Founded: 2015
- Headquarters: Netherlands (NL)
- Pricing tier: Enterprise
- Market side: Employer-side (AEDT)
- Categories: Assessment AI
- Description: Pre-employment assessment suite covering cognitive ability, behavioral, and situational judgement, including the legacy Pymetrics games product.

These are company details, not findings, and carry no weight in the score.

## Head-to-head comparisons

- [Harver vs Fuel50](https://www.hireaiscore.com/compare/fuel50-vs-harver): 56 vs 58
- [Harver vs Bryq](https://www.hireaiscore.com/compare/bryq-vs-harver): 56 vs 57
- [Harver vs Sapia.ai](https://www.hireaiscore.com/compare/harver-vs-sapia-ai): 56 vs 56
- [Harver vs Cornerstone OnDemand](https://www.hireaiscore.com/compare/cornerstone-vs-harver): 56 vs 56
- [Harver vs iMocha](https://www.hireaiscore.com/compare/harver-vs-imocha): 56 vs 56
- [Harver vs HackerRank](https://www.hireaiscore.com/compare/hackerrank-vs-harver): 56 vs 55

## Conflicts of interest

No vendor pays for placement, scoring, or removal. Casework - the consulting firm that operates this directory - provides paid services to some vendors. Any active or recent (within 24 months) commercial relationship is disclosed on the affected vendor profile and the review is reassigned to an independent reviewer.

Casework has no commercial relationship with this vendor.

Full policy: https://www.hireaiscore.com/about

## How to cite

HireAIScore, "Harver" profile, rubric v1.0, last reviewed 2026-11-14. https://www.hireaiscore.com/vendors/harver

Please link the profile URL rather than this file, quote the score with its rubric version and review date, and include the vendor's reply where one is published.
