Bryq

Company overview

Bryq is a talent-intelligence and pre-hire assessment company founded in 2017 (co-founders Markellos Diorinos and Hassan Chahrour, with Guy Krief; George Kalyvas has been CEO since 2025). Its commercial base is Boston, US, with product/R&D operations in Athens, Greece, and the corporate entity behind the product appears on its own trust center as 'Bryq (Idalto Inc.)', with EU affiliates Idalto Ltd. (Cyprus) and Idalto Hellas SPPC (Greece). The platform combines a ~14-minute chatbot-driven assessment (cognitive ability, 16PF/OCEAN-based personality, and 140+ skill tests, plus a newer AI-fluency assessment) with an AI matching engine that builds an Ideal Candidate Profile and scores candidates on a fit tier from 'Exceptional Fit' to 'Poor Fit'. Bryq markets EEOC/OFCCP-validated, blind (name/gender/age/education-stripped) assessment, integrates into existing ATSs, and is a smaller, venture-backed vendor priced for the mid-market.

Regulatory exposure

As an automated tool that scores and ranks candidates, Bryq is squarely in scope of the major AI-hiring regimes: an EU AI Act Annex III high-risk system (it markets to EU employers and runs EU-hosted infrastructure), a NYC Local Law 144 AEDT, and an emerging obligation target under Illinois and Colorado. Its strongest asset is bias-audit transparency: a publicly downloadable NYC LL 144 AEDT bias audit conducted by code4thought (March 2025) discloses a full disparate-impact analysis across 9,846 real candidates with gender, ethnicity, and intersectional scoring-rate and impact-ratio tables (lowest ratio 0.821), on top of a multi-year history of Holistic AI audits since 2023 that remain request-only. Data governance is well documented (SOC 2 Type II and ISO 27001:2022 both audited annually, a public DPA, subprocessor list, and a live Drata continuous-monitoring trust center). Gaps remain typical of the category: no ISO 42001, no published model/system card or Annex IV technical file, no EU AI Act Article 27 FRIA template, and only decision-support-level (not documented override/audit-log) human-oversight detail.

Path to a higher score

Bryq already clears the hardest bar in this category with a genuinely public, full-population code4thought bias audit; the highest-leverage next steps are depth and openness elsewhere. It should publish a formal model/system card and an Annex IV-aligned technical pack (and pursue ISO 42001) to lift Article 11 documentation; keep the code4thought audit on a named yearly cadence and un-gate the prior Holistic AI reports to reach 'two consecutive public audits'; publish an explicit EU AI Act Article 27 FRIA/deployer template rather than relying on a data-protection DPIA procedure; and document concrete in-product recruiter-override controls, audit logs, and per-jurisdiction toggles instead of generic decision-support framing. Adding an AI-specific model-update changelog and incident channel would round out post-market monitoring beyond the existing security-controls dashboard.