- In place today
A person makes the decision
JD Swipe never rejects or ranks a candidate on its own. The summary (meets the JD, close to the JD, or below the JD) describes how the answers compare with the Job Description the recruiter wrote – not the person. It, the per-area results and the observations are inputs for the recruiter. Signals are observations, not verdicts, and every one of them is reviewed by a human. This also keeps decisions outside GDPR Article 22 (no decisions based solely on automated processing).
- In place today
No emotion recognition, no biometrics
Inferring emotions in the workplace, including recruitment, has been banned under the AI Act since February 2025. Voice explanations are assessed only for what the candidate says and for specific technical anomalies (a second voice, speech that sounds read out or synthetic). We do not analyse stress, confidence, personality, accent or any other personal trait, and we do not identify people by their voice.
- In place today
Candidates know what happens
Before the check starts, candidates see who is assessing them, that the questions and the analysis use AI, what is recorded, that the microphone is optional and switched on only when they press it, and that everything is deleted after 90 days.
- In place today
Every question is checked and traceable
Each card is generated for one Job Description and verified three times: answered blind by a second model, reviewed for ambiguity, and checked against official documentation with a source link. Answers, timings, signals and the recruiter's decisions are logged, so any result can be explained afterwards.
- In place today
Data protection by default
Hosting in the EU (Finland), automatic deletion 90 days after a check, backups kept 30 days, a paid AI service that does not use our data to improve its models, and candidate data visible only to the recruiting company.
- Before Dec 2027
Formal compliance for high-risk AI
AI used to evaluate candidates is classified as high-risk. The detailed obligations apply from 2 December 2027. Our plan before that date: a risk-management system, technical documentation, instructions for use for recruiters, a conformity self-assessment, registration in the EU database and CE marking.