AI Notice
This notice explains where AI is used in Vettasy interviews, what it produces, what it never does and how people remain responsible for every decision.
1Purpose
This notice gives Candidates, interviewers and Customers the information that Article 50 of the EU AI Act and Articles 13 and 14 of the GDPR require about AI in Vettasy. It supplements the Privacy Policy.
2Where AI is used
The other Verified Session signals, such as hidden windows and virtual cameras, come from system checks with fixed rules, not from AI models.
3What the AI never does
- Reject, rank or shortlist Candidates.
- Save a score without confirmation by a person.
- Recognise emotions or read facial expressions, tone of voice or personality.
- Speak to Candidates. The synthesized voice in voice translation repeats what a participant said and adds nothing of its own.
- Identify a person by their face. The liveness check keeps no face template.
- Learn from Customer Data. No model is trained on it.
4Human oversight
Every AI draft is labelled as such until an interviewer confirms, edits or discards it. A scorecard is saved only after confirmation, under the interviewer’s name. The decision log records who decided, when and on which evidence, and exports for audit.
Each quote in a draft links to its timecode, so the interviewer can check it against the recording before confirming.
Candidates can ask for human review of any AI-assisted output about them at [email protected] and can dispute any line of the Session Report within 14 days.
5Labelling and transparency
Invitations and the consent screen state that AI is used in the Interview. AI output is labelled in the Application and in the Workspace. The synthesized voice is marked as AI-generated in machine-readable form, and the Application shows when voice translation is on.
6Data used by AI
AI features receive only the content of the Interview and the materials the Customer attaches to the role, such as the plan and the CV. Providers process the data under contracts that forbid training on it. Prompts and outputs on Amazon Bedrock are not stored, and the analysis runs in the Frankfurt region.
7Accuracy and limits
Transcripts and translations can contain errors, especially with background noise, overlapping speech or strong accents. Drafts can miss context that the interviewer noticed. AI output is therefore a starting point that a person checks. The recording remains the reference.
8Legal framework
- Article 5 of the AI Act, Regulation (EU) 2024/1689, has prohibited emotion recognition in the workplace since 2 February 2025. Vettasy does not offer it.
- Article 50 has applied since 2 August 2026. It requires that people are told when they interact with an AI system and that synthesized audio is marked in machine-readable form.
- AI systems used in recruitment are high-risk under Annex III. Following Regulation (EU) 2026/1744, those obligations apply from 2 December 2027. The decision log, records of human oversight, logging and technical documentation are already part of the Service.
- Article 22 of the GDPR restricts decisions based solely on automated processing. Hiring decisions in Vettasy are always made by people.
9Questions
Questions about AI in Vettasy go to [email protected]. Questions about how a specific Customer uses AI output go to that Customer.
