Cette page n’est pas encore disponible en français. La version anglaise est affichée.

AI and data processing

Transparency for the person in the conversation, a bounded configuration, a named owner for every obligation, and evidence that can be examined.

Information tied to a real processing scope.

The legal roles and AI disclosures depend on the actual processing configuration.

  1. Public frameworkHow AI and data processing are understood in the service
  2. Actual projectRoles, providers and controls specified in signed documentation

Document status

Current version · formal approval pending

Version
0.1
Last updated
17 October 2026

This is the current website version. It is not a legal opinion. It describes our standard approach; the agreement with Vee Center Sp. z o.o. governs each deployment.

About this document.

This page holds the full working text. The status block applies to the whole document, and the contents list every section.

Scope of this document

This document describes how AI is used in a Vee operation and how responsibility is allocated. It covers the transparency obligations that apply to an AI system, the configuration behind the service, the human boundary and the evidence a review can examine.

It does not decide the legal classification of a specific deployment. That assessment is made with the client, for the intended purpose, data, components and operating controls.

Transparency: Article 50 disclosure

The transparency obligations in Article 50 of the EU AI Act apply from 2 August 2026. People who interact with an AI system are told that they are interacting with AI.

Where voice or other output is synthetic, that is disclosed as part of the interaction design rather than only in a footer or a help page. The disclosure is a step in the conversation, placed where the person can act on it.

The disclosure does not depend on the customer asking for it, and it is not replaced by a general statement on this website.

Deterministic configuration

The reference configuration favours deterministic dialogue and process rules over uncontrolled generation. The scope of any generative component is bounded, and material autonomous decisioning is excluded from the customer-interaction layer.

Behaviour changes through the approved change process, not implicitly from production traffic. Scenarios, rules and content are versioned, so a reviewer can tell which version produced an outcome.

Closed, approved content catalogue

The operation runs from a closed catalogue of approved voice and message content. The catalogue is versioned and is part of the configuration, not an open prompt surface.

A change is proposed, reviewed, approved and released. The catalogue and its approvals are artefacts a review can examine.

Human oversight and escalation

Processes can include transfer, escalation, callback and stop conditions. Consequential decisions, sensitive requests and exceptions are designed to reach a person, and the handover carries the context the person needs.

Human oversight is a rule in the process rather than an assumption about good intentions. Where the process stops, it stops by design and records why.

High-risk classification by intended purpose

High-risk classification follows the intended purpose of a deployment, not the technology on its own. The same component can sit outside the high-risk category in one use and inside it in another.

Under the current timetable, Annex III obligations apply from 2 December 2027 and Annex I obligations from 2 August 2028. The assessment for a specific deployment is made with the client and recorded before the operation starts.

Provider, deployer, controller and processor

Where Vee builds the AI component and operates it for a client, Vee is the provider of that component and the client is the deployer for its own intended purpose. Where the client supplies a component, the client’s own provider relationship applies. The allocation for a specific deployment is recorded in the deployment documentation, not inferred from the label on the contract.

For personal data processed to deliver the service, the client is the controller and Vee Center Sp. z o.o. is the processor. For personal data processed through this website, Vee S.A. is the controller.

Responsibility allocation

Scroll horizontally

AreaVeeClientWhere it is recorded
AI componentProvider of the reference configuration it builds and operates.Deployer for its own intended purpose; decides the use and the operating context.Deployment documentation.
Personal dataVee Center Sp. z o.o.: processor for service data.Controller for service data.Applicable agreement and data-processing terms.
Regulatory positionSupports the assessment with architecture, configuration and evidence.Remains responsible for its own regulatory obligations.Assessment agreed per deployment.

Processing location

Personal data for the service is processed in the EU/EEA as a rule; any transfer outside the EEA takes place only under Chapter V GDPR and the agreement. Sub-processors and their role are recorded in the deployment documentation and the applicable agreement.

A transfer outside the EU/EEA would require the client’s agreement and an appropriate safeguard. This document does not state that such a transfer takes place.

Incident notification

The applicable OWŚU sets the notification windows. A critical incident is notified within 2 hours; a significant incident is notified within 8 hours after classification.

Classification is made against the criteria agreed for the deployment. Notification does not replace the client’s own obligations to its regulators or its customers.

DORA and financial entities

For a financial entity, responsibility under DORA stays with the client. Vee supports that responsibility with the contractual arrangements, information and evidence the client needs.

Supporting a client is not a compliance status. This page does not claim that Vee or a deployment is compliant with DORA, and the client remains responsible for its own supervisory position.

Evidence a review can examine

  • Architecture diagram and the boundary of each component.
  • Scenario and rule versions.
  • Approved voice and message catalogue.
  • Change approvals.
  • Call and connection logs.
  • Interaction and outcome mapping.
  • Escalation and stop rules.
  • Retention and access rules.
  • Incident and continuity documentation.

The set produced for a deployment is agreed in the deployment documentation. The list is what a review can examine, not a promise that every item exists in every deployment.

The controls behind these statements are described under Trust and governance. The privacy policy covers data roles and rights, and the terms of use cover this website.

Back to top

Walk the AI design before you sign it.

Bring your intended purpose, your classification question and the people who assess it.