picked.
Log inPost a role
Trust · Responsible AI
The principles. The rights. The lines we will not cross.
Six principles. Eight candidate rights. Six things we will not do. The governance that holds it all in place.
Read the candidate rights →See what we don't do
Principles
Six principles. Each operationalised. None decorative.
01
Transparency.
Every candidate knows they are interacting with AI, at every stage, and can ask for a human at any point. The dashboard tells them what was decided, by whom, and why.
02
Explainability.
Every score has a written reason. Every decision has a model version. Every model version has a card.
03
Human-in-the-loop.
Picked scores and ranks. A named human hiring manager makes the final hire decision. No solely-automated significant decision.
04
Candidate consent.
Voice recording, biometric checks, opt-in matching, training-data use: all consented, per session, revocable.
05
No scraping, no model-training on candidate data.
The candidate is not the product. The candidate-owned profile is the only persistent record, and the candidate controls it.
06
Adverse-impact monitoring as a hard stop.
A breach of the 4/5ths rule pauses syndication. We investigate. We resolve. We publish.
Candidate rights
The candidate is not the product.
See every score
Numeric score, written reason, weight. Same view you see. Not a redacted PDF.
Request human review
One click. Any score, any stage. We open a ticket to the hiring manager. 14-day SLA.
Opt out, get human screen
Skip the AI interview. We route the role through human screening instead. Same fee.
Export everything
Transcript, recording, scores, notes. JSON or PDF. Theirs to keep, and re-use.
Delete everything
30-day default retention. Earlier on request. Backups purge within 90 days.
Re-use the vetted profile
Pass an assessment once, carry the result across employers for 12 months.
Know what was recorded
Voice screens and interviews are recorded. The candidate consents per session. No silent recording.
Know who decided
The dashboard tells the candidate "scored by Picked, decided by [hiring manager name, title]".
What we don't do
No scoring on video.
No scoring on faces.
No scoring on names.
No facial analysis
We do not infer emotion, attention, honesty, or anything else from the candidate's face. Video interviews are not part of the V1 product. When video lands with V2, it will be transport, not signal.
No protected-class inference
We do not infer race, gender, age, sexuality, religion, or disability from CVs, voice, or any signal. We monitor outcomes against declared categories, and only with consent.
No model training on your data
Candidate CVs, recordings, assessments and transcripts are never used to train the underlying foundation model. Anthropic operates under a zero-retention agreement for our traffic. OpenAI Whisper runs under the same zero-retention terms.
No personality pseudoscience
No MBTI, no DISC, no five-factor scoring from interview audio. Neuroworx's psychometric IP is rooted in validated workplace ability measures, not pop-personality.
No silent recording
Voice screens and interviews are consented to per session. The candidate sees a counter, can pause, can stop, can re-do.
No re-selling candidate data
Candidates can re-use their vetted profile across employers; that is a feature for them. We do not sell, share, or syndicate candidate data to third parties for marketing or any other purpose.
If something goes wrong
Four paths. All real. All staffed.
01
Candidate review request. Any score, any stage, one form. 14-day SLA. Reviewer is a Picked-trained human with full file access. Original decision paused while review is open.
02
Employer audit request. Any role, any candidate. Per-decision JSON, per-role bias report, model version. Available in the dashboard, exportable.
03
Regulator request. Per-jurisdiction. NYC LL144 audit summary, EU AI Act conformity documentation, EEOC adverse-impact reports. Contact legal@picked.ai.
04
Public complaint. A working appeal path with clear timescales. Not a contact form into a black hole. Email trust@picked.ai; we triage within one business day.
Who decides what changes
Three roles. One change-control process.

Material changes to the production pipeline (a new model version, a new sub-processor, a change to retention or residency) require sign-off from three roles internally: the engineering lead, the trust lead, and the founder. Sign-off is logged, dated, and exposed on the public model cards or sub-processor pages as the change ships.

  • ●Engineering lead: technical change review.
  • ●Trust lead: fairness, privacy, security review.
  • ●Founder: final sign-off on customer-affecting changes.
  • ●All material changes notified to customers 30 days in advance via the sub-processor change feed.
The hiring industry's relationship with AI is going to be defined by the first few products that get it visibly wrong. We would rather be defined by getting it visibly right. Everything on this page exists so that a regulator, a candidate, a journalist, or your own legal team can ask "how does this work, and why should I trust it?" and the answer is one click away.
Related
TRUST · FAIRNESS
Fairness
Per-role bias monitoring. Four-fifths rule. Annual independent audit.
Read →
TRUST · PRIVACY
Privacy
GDPR Article 22. Single-jurisdiction London residency. No training on candidate data.
Read →
TRUST · COMPLIANCE
Compliance
Eight frameworks across four jurisdictions. Per-framework posture.
Read →
Last reviewed · 22 May 2026 · v1.0
Still on the legal review?
We'll send you the file.
Pre-filled vendor security questionnaire (SIG Lite), DPA, sub-processor list, SOC 2 letter, model card, latest bias audit. One zip. Reply within one business day.
Request the legal pack →Email legal@picked.ai ↗
Operated by Neuroworx Ltd · ICO #09910326623
picked.
The AI-native replacement for inbound hiring. For the hiring manager, not HR.
© 2026 Picked · Made in London · Three finalists in forty-eight hours.
Responsible AI · Picked.ai