Score Distribution & Statistical Bias
Histogram of every AI score in the period, with concentration checks, mean and median. If 50%+ of candidates land in one bucket, we flag it as a calibration risk before a regulator does.
Article 15 · AccuracyHiring AI is now Annex III high-risk under the EU AI Act. Most ATSs will charge you for a compliance consultant. We ship the audit pack on every workspace, every day, generated from your live data, no demographic information collected.
Every workspace can generate this report on demand. The same numbers regulators see, the same numbers your hiring team sees.
Histogram of every AI score in the period, with concentration checks, mean and median. If 50%+ of candidates land in one bucket, we flag it as a calibration risk before a regulator does.
Article 15 · AccuracyWe never collect demographic data. Instead, we measure score variation across name-origin patterns, education tier, location and career-gap presence. Any dimension drifting >10% from baseline is amber-flagged.
Article 10 · Bias mitigation% of AI decisions reviewed by a human, % of AI rejections rescued by reviewers, median time-to-review, and decisions made without human review. Article 14 compliance is computed live.
Article 14 · OversightGenerated from your workspace config: primary model, provider, purpose, EU AI Act classification, GDPR legal basis, retention, right-to-explanation policy, erasure requests handled.
Article 13 · TransparencyWe don't trust dashboards either. Every metric in TenPerZent flows through canonical SQL functions, audited against the raw event log. If two independent paths disagree, we alert ourselves before you ever see a wrong number.
Two-page audit pack with synthetic data. Real format. Yours is generated live from your workspace.
Sample uses synthetic data. Not legal advice.
Based on each vendor's public documentation, security pages, and product capabilities as of 2026.
| Capability | TenPerZent | Workable | Ashby | TeamTailor | Manatal | Lever |
|---|---|---|---|---|---|---|
| Score-distribution audit (statistical bias) | ||||||
| Proxy fairness without collecting demographics | ||||||
| Article 14 human-oversight metrics | ||||||
| Auto-generated model card | ||||||
| Right-to-explanation per candidate | ||||||
| Drift detection on KPIs | ||||||
| Reconciled / audited metrics | ||||||
| Downloadable audit pack on demand |
Comparison based on publicly available product documentation. Contact us if your vendor has shipped a comparable feature, we'll update.
Every third party that processes any candidate or workspace data, what they do, and where they sit. Updated whenever it changes — last reviewed April 2026.
| Sub-processor | Purpose | Region | DPA |
|---|---|---|---|
| Supabase (via Lovable Cloud) | Database, auth, file storage, edge compute | EU — Frankfurt | Yes |
| OpenAI | CV parsing, scoring, JD enhancement | EU API endpoint, zero-day data retention | Yes |
| Google Gemini | Backup AI provider, multilingual scoring | EU API endpoint, no training on data | Yes |
| Stripe | Subscription billing & top-ups | EU + global (PCI-DSS L1) | Yes |
| Resend | Transactional & auth email delivery | EU | Yes |
| Cloudflare | DNS, edge caching, DDoS protection | Global edge, EU-anchored | Yes |
| Sentry (optional) | Error monitoring (no candidate data) | EU | Yes |
It's built into every TenPerZent workspace. Start free, ship the audit pack on day one.
Informational, not legal advice. EU AI Act and GDPR compliance is a shared responsibility between provider and deployer.