Criteria

Company overview

Criteria Corp, founded in 2006 and headquartered in Los Angeles, is a pre-employment assessment vendor that publishes cognitive aptitude (e.g., CCAT), personality, emotional-intelligence and skills tests backed by industrial/organizational psychology research, alongside structured-interview and AI-assisted interview-intelligence products. It serves thousands of customers across 50+ states and 60+ countries, and tests its assessments on job-relevant data covering more than 10 million candidates annually. Its public 'science' and 'security and compliance' pages position the company around validity, fairness and EEOC/SIOP/APA testing standards rather than around a dedicated regulatory-compliance program.

Regulatory exposure

Criteria's assessment scores 'substantially assist or replace' discretionary hiring decisions, placing its tools squarely within scope of NYC Local Law 144 (AEDT), the EU AI Act's Annex III high-risk employment category, and US state AI-hiring laws (Illinois, Colorado). The company states it commissions independent third-party bias and disparate-impact audits and benchmarks against NIST, IEEE, ISO and the EU AI Expert Group, and it holds ISO 42001:2023, ISO 27001:2022 and SOC 2 Type II. However, the actual bias-audit summaries are not openly downloadable (gated to clients), and the company publishes no EU AI Act deployer guidance, no FRIA template, and no explicit Local Law 144 employer-compliance pack; its CEO has publicly criticized the NYC AEDT rule rather than offering deployers a compliance path.

Path to a higher score

Criteria already has unusually strong foundations for an assessment vendor (ISO 42001, ISO 27001, SOC 2 Type II, a public AI-principles page and a DPA). To raise its score it should publish its third-party bias/disparate-impact audit summaries as openly downloadable reports naming the auditor and date (ideally two consecutive years), release an EU AI Act deployer pack with a FRIA template and Article 26/27 guidance plus a Local Law 144 employer kit, publish an AI/model card or system card and instructions-for-use leveraging its ISO 42001 program, and stand up a public status/incident or model-update channel to evidence continuous post-market monitoring.