AI transparency

Updated 17 October 2026

HR Copilot uses AI to help employers hire fairly and quickly. This page explains where AI is used, what it's told to ignore, what it never does, and the part people play.

1. Where HR Copilot uses AI

AI helps at each step of hiring, from reading CVs to interviews, but it gives advice, not decisions.

  • CV scan: reads a CV and the candidate's answers against what the job needs, gives a score, and lists strengths, gaps and what's missing.
  • AI interview: the AI Recruiter interviews candidates by voice, asks the employer's questions and follows up within the limits the employer sets. It always introduces itself as an AI Recruiter.
  • Interview scorecard: scores the interview against the employer's criteria and points to the moments worth hearing.
  • Assignments: helps employers write assignments and scores candidates' submissions against the employer's criteria.
  • Integrity checks: looks for signs that someone else is helping during an AI interview (see section 5).
  • Everyday help for recruiters: writing and rewording, translation, speech to text and the Calendar assistant.

2. A person decides

AI scores and summaries help the employer's team. The team decides who moves forward and who is hired.

  • The AI suggests a next step. It's only a suggestion.
  • Every AI score, and every change a person makes to one, is recorded with the time.
  • One exception an employer can choose: automatic rejection. An employer can enable it for a job, so a CV that scores under a level the employer sets is rejected after a short wait, during which the team is told and can stop it. A candidate rejected this way can reply to the rejection email to ask for a person to review the decision.

3. What the AI is told to ignore

The AI is instructed to judge only job-related evidence, and never personal traits.

When scoring, the AI is instructed never to consider, guess at or mention a candidate's:

  • name or photo
  • age or date of birth
  • gender
  • nationality, ethnicity or where they're from
  • religion
  • marital or family status, or pregnancy
  • disability or health
  • appearance or accent

This holds even when a CV states them. In addition:

  • Language is judged only when the job lists it as a requirement, and then only as that requirement. Answers in Arabic and English count equally.
  • Right to work, location and similar requirements count only when the job lists them.
  • Gaps in employment aren't treated as weaknesses, unless the job requires recent experience in something.
  • Attempts to game the AI, such as hidden instructions in a CV, are ignored and flagged for the team.

We design HR Copilot to support fair, consistent hiring, and we keep checking how it scores. We don't claim the AI is free of all bias; that is why a person decides.

4. What the AI never does

It scores what a candidate says, never how they look or sound.

  • It doesn't score tone of voice, facial expressions or emotions.
  • It doesn't use a candidate's face to recognise who they are or to judge their appearance.
  • We don't build or keep voiceprints or face templates, and we never use someone's voice or face to work out who they are.
  • It doesn't penalise accents, filler words or transcription mistakes.
  • It doesn't make the final hiring decision.

5. Integrity checks

During an AI interview, we look for signs that someone else is helping, and a person reviews anything we find.

We check for signs such as another voice or another person on camera, using the interview recording and still images from the camera. Anything found is flagged for the employer's team to review. A flag never rejects anyone on its own.

6. Learning from a team's score changes

A team can let HR Copilot learn from the reasons it gives when changing a score, within its own workspace only.

When someone on the team changes an AI score and says why, HR Copilot can turn the reason into a short, general guide for later scores, such as "count hands-on IFRS reporting as strong evidence for finance roles", so scores follow the team's standards.

  • Each reason is checked for fairness first. A reason that relies on a personal trait, openly or indirectly, still counts for that one candidate as the person decided, but teaches nothing, and the team's owners and admins are told.
  • Guides contain no names or details about any candidate.
  • They're used only in that team's own workspace, and never leave it.
  • Owners and admins can read, rewrite, disable or delete any guide, or disable the whole feature.

7. Where the AI runs

Our AI features run on AI services from specialist providers, under contracts that protect the information.

The AI services we use process text and audio in the United States, under data agreements described in our privacy policy. They're used only to provide HR Copilot's features.

8. Candidates' rights

Candidates can ask for a person to review an automatic decision, and to see, correct or delete their information.

Candidates can ask the employer they applied to, or write to privacy@hrcopilot.ae and we'll help. See our privacy policy and the candidate notice shown when applying.

9. The limits of AI

AI can be wrong. Employers should review what it suggests.

AI can misread a CV, mishear an answer or miss context a person would see. That's why employers review AI scores and summaries, why every score can be changed, and why a person makes the decisions.

10. Questions

Write to privacy@hrcopilot.ae about how HR Copilot uses AI.