Defensible by artifact,
not by assertion.
AI hiring tools are getting audited — by the FTC, by the EEOC, by NYC, by GDPR regulators, by procurement. Most of them ship without the receipts. We’re building toward leading with them. Below is our honest posture: what’s built, what’s in progress, and what we won’t claim until it’s real.
What we operate under — and how literally to read each line.
We treat candidate background screening as an FCRA-class workflow. Pre-adverse-action notice + dispute window are part of the design; the screening integration that fully automates this is queued for Phase 3 of the build. Until then, the FCRA workflow runs manually with the same artifact trail.
Our scoring model is designed for fairness monitoring — protected-class log fields are in the schema and the impact-ratio reporting flow is ready to populate as cohort sizes grow. We don’t publish bias-audit numbers yet because the sample sizes per cohort are small enough that the headline stats would mislead more than inform.
The pipeline produces the artifacts an LL 144 audit asks for — protected-class breakdowns, candidate notification copy, AEDT documentation. We’ll formalise the published audit once we have a buyer in scope and the cohort sizes are statistically meaningful.
Right to access, right to delete, right to object to automated decision-making. Our DSR workflow is human-handled today and will move to a self-serve endpoint as EU candidate volume grows. EU residency available on request — Supabase region targeting is configurable.
Income disclaimer + Meta Special Ad Category.
No earnings claim — about the platform, a placement, a partner track, or a graduate — ships without sample-size context, time window, and the explicit statement that individual results vary. “Up to” / “as much as” framing is banned in cert criteria. Outcome stories must be traceable to a verified results-gate proof.
cyborgtalent ads run under the Employment Special Ad Category on Meta — restricted targeting, no protected-class proxies, no narrow geo + age combinations that imply discrimination. Our paid acquisition team operates under that constraint by default. (For the avoidance of doubt: this site itself is not yet running Meta ads.)
What every placement ships with.
Override Instinct · Brief Discipline · End-to-End Ownership · Honest Uncertainty · Pattern Recognition · Composure. Each scored with verbatim evidence from the application + sim.
Voice interview audio + transcript + post-call analysis. Signed-URL access (7-day default), longer windows on request.
Every Tier 2 + Tier 3 deliverable, with the AI rubric score AND the manager / peer review notes attached.
The Cyborg Trait composite, weighted 0–100, evolving across every cert event. Not a one-shot number — a signal that compounds.
Once placed, downstream client outcomes flow back. The audit packet updates with real performance data over time.
Tier 1 scoring impact ratios across protected classes, refreshed as cohort sizes reach the threshold for meaningful reporting. Available to your compliance team on request.
We didn’t add compliance last quarter. We’re building toward leading with it.
A few things adjacent platforms market loudly that we won’t say until they’re fully shipped:
- SOC 2 Type II. Not certified. We operate against the Trust Services Criteria internally; control inventory is available under NDA. Formal audit is on the 2026 roadmap.
- Published bias-audit numbers. Cohort sizes per role are still small enough that the headline numbers would mislead. Aggregate ratios are shared with active buyers under NDA.
- FCRA-automated dispute flow. Manual today. Fully-automated dispute + re-review workflow is queued behind the background-check integration.
- Income claims tied to placements. We don’t. We won’t until the results gate is live and the cohort sample is statistically meaningful.
Where the data lives. How long it stays. Who can see it.
Supabase hosted with regional targeting. EU residency available. Audio recordings on signed-URL storage with TTL-based expiration.
Application + Tier 1 data: 24 months default. Cert artifacts: held while pool membership is active + 12 months. Sim recordings: 7-day signed URLs by default. All overridable per contract.
Row-level security on every table. Cross-client data isolation by design. Buyer access scoped to their own placements. Operator access logged and reviewed.
What buyers ask first.
Are you SOC 2 certified?
What happens when a candidate disputes their score?
Do you publish your bias-audit results?
Can I get the raw audit packet for one of your placements?
What about EU candidates?
Where can I see your terms?
Bring the audit. We’ll bring the artifacts.
If your procurement team needs to sign off, send them this page and we’ll handle the rest.