Artificial intelligence and transparency
Updated 24 September 2026
Website and blog content
Blog texts are developed with substantial assistance from artificial intelligence, based on VoxVent design notes and conversations. This disclosure does not attest to an individual human review of every article. Examples are illustrative, not reports of real events. Features described should be read alongside the development and testing status stated on the website.
Interpreters and transcription: two different functions
The voices on interpretation channels belong to the connected interpreters. VoxVent does not automatically generate their translations. Transcription has been removed from the new Mac 0.21.1 test package and deferred to future development. The caption guidance below applies to earlier prototypes that include this feature, not to a feature available in the new production app.
How transcription works
In earlier prototypes with transcription, Apple Speech recognition is requested on the production device. If on-device recognition is unavailable for the selected language, the feature must not start with an alternative cloud service. The production app can display text on a monitor or send it over the production network using NDI. The portal does not automatically detect when captions are enabled in the Mac app.
The website and portal do not offer an AI chatbot. Cloud AI translation of captions is a future development, not an active feature of this version.
Limitations and human oversight
Automatic text may contain errors, omissions or words that were not spoken, particularly with noise, overlapping speech, accents, proper names and numbers. It is not a certified transcript and must not be used alone for medical, legal, safety or individual-rights decisions. Refer to the speaker and interpreters to confirm content.
The production operator chooses the source and language, monitors the text and can stop transcription. To report an error during an event, contact the organiser; for product issues, write to andrea@receventi.it, stating the version, language and circumstances, without sending unnecessary audio or personal data.
Information for the audience
When automatic captions are displayed, the organiser must clearly disclose their origin before or at the first display. Suggested wording: “Captions automatically generated using AI. They may contain errors.” Keep the notice legible on the screen or video composition displaying captions, including when the audience does not visit the portal.
Before using the feature at an event
- Identify who monitors captions and who handles reports.
- Test the language, names, numbers and audio under event conditions. Check that the operator knows how to stop the output and flag an error.
- Inform speakers and participants about processing, text destinations and any retention; also see the privacy notice.
- Display the notice on the actual screen seen by the audience. An information page alone does not replace the contextual notice.
- Before publishing the text as verified content, assign substantive review to a competent person and establish editorial responsibility.
European AI Act
VoxVent measures are based on analysing the actual use of features and the roles of the supplier and organiser. User transparency, assessment of any technical content marking and operator preparation require separate actions. This page is not a compliance certification and does not attest that operators have completed training.
No emotion recognition, biometric identification, assessment of individuals or automated decisions about their rights are planned. New engines or uses require a new assessment before activation.
References: AI Act, consolidated text; European Commission: transparency; European Commission: AI literacy.