HiPeople

Company overview

HiPeople GmbH was founded in Berlin in 2019 by Jakob Gillmann and Sebastian Schüller and is registered at Zionskirchstraße 73a, 10119 Berlin (HRB 211492 B). The company raised roughly $6.8M from Moonfire, Capnamic Ventures, Cherry Ventures and Mediahuis. Its platform spans four modules: a talent assessment library of 400+ skills and personality tests, automated reference checking with fraud/relationship verification, AI resume and application screening that produces a match score against customer-defined criteria, and AI screening calls across 113 languages. Customers named on its trust center include Fujifilm, Douglas, OLX Group, DAZN, Celonis, Zapier and Grammarly. Entry pricing starts around $70/month with a credit-based model layered on top.

Regulatory exposure

As a Berlin-headquartered provider whose tools screen and rank candidates, HiPeople sits squarely in Annex III employment high-risk territory under the EU AI Act and carries provider-side obligations, with GDPR Article 22 exposure on automated decision-making. Its US customer base pulls in NYC Local Law 144 AEDT duties, plus EEOC/Title VII and the Illinois and Colorado regimes. HiPeople is unusually strong on the bias-audit dimension: Warden AI runs continuous monthly third-party auditing of its AI Application Screening system, with a publicly viewable dashboard (latest audit 9 July 2026, 19,092 samples) covering sex, race/ethnicity, intersectional sex-by-race, age, disability, religion and sexual orientation, plus dedicated NYC LL 144 and EU AI Act bias reports carrying published selection rates and impact ratios. The important caveat is scope: that assurance covers the application-screening product only, not the assessment library, reference checks, or AI screening calls. Elsewhere the picture thins — the SafeBase AI trust center structures topics against the AI Act but gates or client-renders the substance, there is no ISO 42001, no model or system card, no published validity evidence behind the assessments, and nothing on Article 27 FRIA support for deployers.

Path to a higher score

The highest-leverage move is extending Warden's continuous auditing beyond AI Application Screening to the assessment library, reference checks and AI screening calls, since those are the products the primary category rests on and they currently carry no published fairness evidence. Publishing an Article 11-style technical pack or system card per AI system — intended purpose, model and data provenance, known limitations, performance metrics — and ungating the 'How to Recognize AI in the Product' and 'AI Training Data' items would lift technical documentation materially. Adding an explicit deployer-obligations page with an Article 27 FRIA template and DPIA mapping would move FRIA support off the floor. Pursuing ISO 42001 alongside the existing SOC 2 Type II, publishing the psychometric validity and adverse-impact evidence gestured at on the /science page, and adding a real AI model-update changelog plus documented override and audit-log controls in the help center would round out the remaining gaps.