AI and the AI Act
At a glance
- In White Mirror, on the candidate’s side and on the company’s, you are talking to AI, not a person.
- AI only makes suggestions. The candidate decides what goes into their avatar and what the company sees.
- AI does not assess people, rank candidates or decide who works with whom.
- We are conducting a self-assessment under the EU AI Act and will publish the findings after legal review.
1Where AI is used in the service
White Mirror for candidates
An in-depth interview with the candidate. The model asks questions and suggests avatar entries — experience, skills, Talents, potential — and points on Core Stats scales, each with references to what the candidate said.
White Mirror for companies
A conversation with a company about an opportunity. The model helps draft a description of the opportunity and its conditions; the company edits and publishes it itself.
Card hint
When a company requests a card, the model suggests which avatar entries may be useful for that opportunity and explains why. The candidate selects and deselects the checkboxes.
Dictation
Speech recognition starts when you click. The text appears in a draft first; you send it yourself.
The model provider is OpenAI. AI-generated text in the service is always presented as a suggestion, rather than a final result.
2What AI doesn't do
- does not grade, score or rank people;
- does not choose which candidates to show a company or rank them — Radar calculates how the stated information matches the opportunity using transparent rules, without a model;
- does not fill in Core Stats on its own — White Mirror may suggest a point on a scale with its reason and your words; only the candidate accepts or declines it;
- does not recognise emotions or analyse voice, facial features or biometric data;
- does not decide who works with whom, whom to turn down or on what conditions.
3Human oversight
The candidate accepts, edits or rejects each suggestion from the model; rejected suggestions are not offered again. The card is sent to the company only after the candidate has reviewed and approved it. The company decides whether to reply to the candidate. The model's responses may contain errors — these can be discussed and corrected in the conversation.
4EU AI Regulation (AI Act)
Regulation (EU) 2024/1689 places AI systems used to select people for work — employment and self-employment alike — among areas subject to heightened scrutiny. We are conducting our own assessment of how the Regulation's requirements apply to CandidDate's features, taking into account the changes introduced by the Digital Omnibus package. This page meets the transparency requirement: you know when you are talking to AI and what it does.
This page is not an EU declaration of conformity or a certificate. We will publish the self-assessment results after review by an independent lawyer.
5Questions
Send questions about AI use to compliance@candiddate.eu.
Terms · Privacy · Security · Principles · Sign in