Our position on the EU AI Act
In short. On Your Assessor, AI helps a company write the text of a questionnaire. It never sees the answers or results, and it does not score, rank or decide about anyone. In our assessment this is not a high-risk AI system under Regulation (EU) 2024/1689 (the AI Act). The company reviews the drafts, and nothing reaches participants until it publishes the questionnaire. In the coach practice module, AI can also draft a summary of a session from the coach's notes, only when the coach has switched it on and the client has consented; the coach reviews the draft, and it is not used to make decisions about the client (section 7).
1. What AI does on Your Assessor
The questionnaire builder in the company workspace can draft, translate and check the text of questions. It uses Claude, a general-purpose AI model of Anthropic, through Anthropic's API; the provider is Anthropic Ireland, Limited, as stated in our privacy notice. The model receives only the description the company writes and the text of the questionnaire.
2. What AI does not do
- It never receives participants' answers or campaign results. An automated check verifies this before every release.
- It does not score, rank, profile or make decisions about people. Scores are calculated from the answers with fixed arithmetic rules, without AI.
- No emotion recognition and no biometric data. None of the practices prohibited by Article 5 of the AI Act is used; in particular, there is no emotion recognition at the workplace (Article 5(1)(f)).
3. Classification: not a high-risk AI system
Annex III, point 4 of the AI Act lists as high-risk the AI systems used in employment (a) for the recruitment or selection of people, and (b) to make decisions affecting work-related relationships, promotion or termination, to allocate tasks based on individual behaviour or personal traits, or to monitor and evaluate the performance and behaviour of people. The builder does none of these: it writes question text, which the company then reviews and edits before it publishes the questionnaire. It does not evaluate the people who answer.
We do not offer a recruitment mode. Customers must not use the AI features to evaluate or select individuals, and every decision about a person must be taken by a human (see section 12 of our Terms of Service). Should we ever add a function that could fall under Annex III, we would assess it before release and update this page.
4. Roles under the AI Act
- Anthropic is the provider of the general-purpose AI model. The obligations of Chapter V of the AI Act for such models are Anthropic's.
- Y2Y Hungary Kft. integrates the model into its service and makes the builder available as an AI system within Your Assessor.
- Companies that use the builder act as deployers of it in their own organisation.
5. Transparency
- The builder is labelled as AI in the application, and its output is marked as an AI draft, a suggestion to be checked.
- The company reviews the draft and decides which questions to keep; nothing reaches participants until the company itself publishes the questionnaire.
- Participants answer a questionnaire that the company has written or approved. They do not interact with an AI, so the chatbot disclosure of Article 50(1) does not arise.
- Article 50(2): question text that the builder writes or translates carries a machine-readable AI marker in the questionnaire's data, and an AI label in the editor, so that it is detectable as generated by AI.
6. AI literacy (Article 4)
We make sure that the Y2Y staff who build and run the AI features understand how they work and where they can fail. Companies using the builder should make sure their own users know that AI drafts can be wrong or unsuitable and must be checked.
7. Coach module
The coach practice module can draft a summary of one session from its notes, using Claude through the Anthropic API (Anthropic Ireland, Limited). It runs only when the coach has switched AI summaries on and the client has given the separate consent for them, and the coach starts each summary. The result is an unpublished draft that the coach reviews and edits; nothing is published or sent to the client automatically, and the draft is not used to make decisions about the client. Article 50(2): a draft is stored with a machine-readable AI marker (author type "AI draft"), which stays on the note after the coach publishes it. In our assessment this is a limited-risk use, not a high-risk AI system.
8. Timeline and updates
The AI Act's prohibitions and the AI literacy duty apply since 2 February 2025, the obligations for general-purpose AI models since 2 August 2025, and the transparency obligations of Article 50 since 2 August 2026. Regulation (EU) 2026/1744 (the Digital Omnibus on AI, in force since 27 July 2026) moved the obligations for high-risk systems listed in Annex III to 2 December 2027 and turned the AI literacy duty of Article 4 into a duty to take measures that support AI literacy; neither changes our position above. We follow the guidance of the European Commission's AI Office and will update this page when the law, the guidance or our product changes. Questions: info@y2y.hu.
TermsImprintPrivacySub-processorsDPA for companiesAI ActSecurity