AI recruiting and compliance
Using AI in hiring means you should be able to explain the process to a candidate and to your own team. This page is a practical checklist for that explanation. It is not legal advice, and it is not a substitute for counsel in your jurisdiction.
Say what the system does
Write down which steps are automated. Sourcing, screening, and scheduling are different decisions from “the AI hired them.” If a person makes the offer, say that. If a score alone rejects someone, say that too. Vague claims are what create risk.
Keep the scorecard next to the result. A note that says “advanced because they have run on-call for a similar API” can be reviewed. A hidden rank cannot.
Keep a person on the decision
A human should be able to overturn a screen and should look at disagreements, not only at the people the model liked. Research on AI hiring bias is clear that people often copy a model’s recommendation even when it is skewed. The oversight has to be real: a rubric, notes, and a review of who was rejected.
Employers AI publishes work on human oversight and on how we monitor bias in hiring models. Read those before you treat a score as neutral.
Data you do not need
Do not feed the screen photos, names used as a proxy, or anything you would refuse to ask in an interview. Collect what the role requires. Retention should match your privacy policy: keep the record of a hiring decision, and do not keep source material you have no reason to store.
What to tell candidates
If an automated step evaluated them, a plain sentence in the application flow is appropriate: software screens applications against the job requirements, and a person reviews the outcome before an offer. Point them at your privacy policy for what you store. If your region requires a specific notice or opt-out, follow that rule. This page cannot list those duties for you.
Common questions
- Is this legal advice about AI hiring?
- No. It is a product and process checklist. Talk to counsel about notice, bias audits, and retention rules where you hire.
- What should we document when AI screens candidates?
- The requirements you scored, the notes the screen produced, and the person who accepted or overturned the result.
- Can a candidate ask for a human review?
- You should be able to provide one for any decision you would defend. Build that into the process before you need it.