Bonmoniq legal
AI transparency notice (Article 50 of the AI Act)
- Last updated
- 2026-10-25
This English version is provided for information as a free translation. The French version is the official legal version. If there is any conflict, the French version prevails.
This notice describes the transparency measures that Bonmoniq SASU applies to the AI features of the Service, in particular the AI voice agent, and the requirements of Article 50 of Regulation (EU) 2024/1689 on artificial intelligence (the "AI Act"), which apply from 2 August 2026, that are relevant to those features. Each section states which measure is in place; a point marked as to be completed is not yet stated as implemented.
It supplements the Privacy Policy and the Data Processing Agreement, to which it refers for the processing of personal data. A feature described here is commercially available only when it is offered in the Service.
1. Scope
The Service offers AI assistance features (conversation assistance, summaries) and, with the telephone channel included in the company subscription, an AI voice agent that answers inbound calls received on the Customer's numbers. The AI voice agent does not place automated outbound calls to third parties. When the Customer turns it on, a caller may interact with a conversational AI system rather than with a natural person.
Transcripts and summaries are produced automatically; summaries are marked in the Service as written by AI. By default, an AI agent works in Suggest mode: a person approves each proposed action. Where the Customer chooses Automatic mode, the agent performs, without prior human approval, the actions the Customer has authorized (for example creating a task, a calendar entry or a note); each action is recorded and attributed to the agent. Where this function is enabled for a Canadian line, the AI voice agent may send to a Canadian number a short confirmation text message with fixed wording, which states that it is an automated message from an AI assistant.
Live call captions. Live call captions start automatically with the call in the Telephony module. Neither the Customer nor an employee can turn them off; there is no separate manual start or captions toggle. Speech recognition operated by Bonmoniq turns both parties' speech into text for live display and uses no external speech-recognition or AI provider. Captioning does not itself record the call or create a saved caption transcript. The initial call announcement informs participants about captions together with the recording announcement; there is no separate mid-call captions announcement. Recording controls remain separate: stopping a recording, including by pressing 9 where offered, does not stop captions. Captions may contain errors, change while provisional or omit words. They assist with following the call, are not a certified record and do not make automated decisions about a person. Human speech captioned in this way remains human speech, not AI-generated speech. In an AI-agent or call-flow test, captions may display caller text already recognized for the agent and the exact agent response text without transcribing synthetic audio again.
If a participant objects to captions, or required consent is absent or withdrawn, the Customer must end the call rather than continue that processing. Captions cannot be disabled while continuing the call. Ending the call does not itself delete any separate recording or other retained content; applicable requests follow the Privacy Policy and DPA. This policy does not assert that the Service automatically detects an objection or consent status.
2. Informing the caller (Article 50(1))
At the start of every call handled by the AI voice agent, a notice tells the caller that they are interacting with an artificial intelligence. This notice is played before any answer from the agent and cannot be interrupted; if it cannot be played, the agent does not answer. The agent is also configured to confirm that it is an AI if the caller asks.
3. AI-generated audio content (Article 50(2))
The AI voice agent's answers are synthetic voices. The notice described in section 2 tells the caller that the voice is artificial. That spoken notice is separate from the machine-readable marking described below and does not replace it.
Marking. As of the date of this notice, synthetic audio newly generated by the Service with a Standard or Premium voice carries a machine-readable audio watermark identifying it as artificially generated. The watermark is embedded in the audio signal by a watermarking model operated by Bonmoniq on its own infrastructure, after synthesis, including for Premium audio returned by Cartesia. It covers in particular the agent's answers during a call, its spoken notices, voice previews, and greetings and messages generated from text in the Service. During a call, if the watermark cannot be applied, the agent stops the affected synthetic speech rather than playing it unmarked.
What is not marked. The watermark does not apply retroactively and is not added to human speech:
- audio generated before the date of this notice, including greetings or messages already stored, is not guaranteed to carry the watermark unless it is generated again;
- human voices, in particular the caller's voice, the voice of a member of the Customer's team and audio files uploaded by the Customer, are not marked as artificially generated;
- a call recording that contains both human speech and synthetic speech is not marked as entirely AI-generated: the synthetic segments it contains may carry the watermark, and recordings are not rewritten.
Limits. Detection of the watermark may weaken or fail on short excerpts, or after editing, added noise, compression, re-recording or other transformations, including some telephone network conversions; detection may also occasionally give an incorrect result. Bonmoniq does not claim that the watermark is detected in every case. The watermark indicates the presence of this marking only: it is not an electronic signature and does not prove the authenticity of a recording, the identity of a speaker, or the origin or history of a file. Its presence or absence does not by itself establish compliance with any legal obligation, by Bonmoniq or by the Customer.
4. Where the AI runs
Reasoning. Understanding and generating the answers of the AI voice agent and of the other AI features run on models operated by Bonmoniq. Conversation content is not sent to a third-party AI provider for this reasoning.
Standard voices. By default, the agent speaks with a Standard voice, synthesized by Bonmoniq without calling an external provider.
Premium voices. Premium voices are reserved for Workspaces with a paid company subscription. Where such a Customer explicitly selects a Premium voice, including to listen to it before choosing it, the text to be spoken is synthesized by Cartesia, currently the only Premium provider. The provider receives the text to be spoken, which may contain personal information from the conversation, together with the chosen voice, language and settings, and returns the generated audio; it receives neither the caller's voice nor the call transcript. The agent's notices are spoken with that voice; a notice in a language that voice does not speak is spoken by another Cartesia voice, failing that by a Standard voice. A Standard voice, and its notices, never call an external provider. Premium voices are available only when Bonmoniq has enabled them for the platform; otherwise, the agent uses a Standard voice. Retention, transfer and training conditions at the Premium provider are described in the Privacy Policy.
5. Reaching a person
A caller who wishes to speak to a natural person may ask during the call. The request does not guarantee an immediate transfer to a person: depending on the configuration chosen by the Customer and the availability of its team, the caller may leave a message or be called back, and the call is passed to a person only where the Customer's call flow provides for it.
6. No automated decisions
The AI voice agent does not take decisions producing legal effects on the caller, or similarly significantly affecting the caller, based solely on automated processing (Article 22 GDPR).
7. Customer responsibilities
A Customer that turns on the AI voice agent remains responsible for its use, including:
- the lawfulness of the purposes for which it deploys it;
- informing callers as required by applicable rules, in addition to the notice played by the Service;
- complying with the rules applicable to call recording and transcription (see the Privacy Policy);
- not configuring the agent to deny that it is an AI.
8. Data
The processing of personal data in connection with the AI voice agent and live call captions, including recording and transcription, retention periods, legal bases and data subject rights, is described in the Privacy Policy and governed by the Data Processing Agreement.
9. Updates
Bonmoniq updates this notice when the Service or applicable obligations change.
10. Contact
AI and notice questions: legal@bonmoniq.com Personal data questions: privacy@bonmoniq.com
Last updated: 25 October 2026 For any question regarding this document, contact us at legal@bonmoniq.com.