AI transparency notice
Last updated:
We develop AI voice agents that speak with people on the phone on behalf of our customers. This notice explains how we make the use of artificial intelligence transparent, in line with Article 50 of Regulation (EU) 2024/1689 (the Artificial Intelligence Act, "AI Act").
Legal framework
Article 50 of the AI Act sets transparency obligations for providers and deployers of certain AI systems, including systems that interact directly with people and systems that generate synthetic audio. It applies from 2 August 2026.
Within the meaning of the AI Act, we act as the provider of our voice agents. The organisation that uses a voice agent in its own name — usually our customer — is the deployer.
AI disclosure at the start of every call
Our voice agents are designed so that the person on the line is informed, at the start of each call, that they are speaking with an AI system. The disclosure is part of the agent's greeting by default, in clear and understandable language, at the latest at the first interaction (Art. 50(1) and (5) AI Act).
Marking of synthetic audio
The voices of our agents are generated by our own speech synthesis. Audio generated by our systems is marked in a machine-readable format and detectable as artificially generated, as far as technically feasible and in line with the generally acknowledged state of the art (Art. 50(2) AI Act).
Voice samples on this website
The voice samples on this website are AI-generated. They were produced by our speech synthesis and are not recordings of a human speaker; they are labelled as such next to each player. The dialogues in them come from test calls with invented customers.
Human oversight and handover to a person
Callers can ask to speak to a person. Where the deployer has set up a handover destination, the agent transfers the call to a human operator when the caller asks for it or when the conversation requires it. Deployers define the scenarios, the information the agent may give and the escalation rules, and they can review calls in the console within the limits of applicable law.
What our voice agents are not designed for
Our voice agents are not designed for emotion recognition or biometric categorisation.
Obligations of deployers
Organisations that use our voice agents are responsible for their own use of them. In particular they should:
- keep the AI disclosure in the greeting of every agent and not present the agent as a human;
- inform callers about call recording and the processing of their personal data as required by the GDPR and national law;
- disclose content that constitutes a deep fake within the meaning of Art. 50(4) AI Act, if they create any;
- assess whether their specific use case triggers further obligations under the AI Act or sector-specific law.
We support our customers with documentation, configurable greetings and the information they need to meet these obligations.
Contact
Questions about this notice or about the use of AI in our products: office@aifortis.com.