AUVY GmbH · Am Haag 8 · 82166 Gräfelfing · HRB 311039 (AG Munich)
Contact: legal@auvy.ai
Version: 1.0 · Last updated: 2026-10-31
AUVY is an AI assistant for work, so what AUVY writes, says and does comes from AI models. This page explains where you meet AI in AUVY, how AUVY tells people they are dealing with an AI, how it marks what it makes, and what stays with you as a business customer. It follows Article 50 of the EU AI Act (Regulation (EU) 2024/1689). AUVY GmbH is the provider of the systems described here.
Where AUVY uses AI
The assistant answers and acts on the work you give it access to, such as files, mail, calendars, chats and meetings. Bots and processes run recurring work on their own, with the access their owner granted. In a call, AUVY can talk with you. A meeting bot joins online meetings to record, transcribe and write notes. AUVY also reads the sources you connect to point out what changed.
None of these is a high-risk AI system under the AI Act, and none of them makes decisions about people.
How AUVY says it is an AI
In the AUVY app you are talking to AUVY, and it says it is an AI whenever someone asks. Outside the app, AUVY makes it clear as follows:
- Mail AUVY sends on its own, without a person approving it first, says so once. The first mail in a thread, or the first to an address, signs off with the sender's name and "AI assistant" (in German "KI-Assistenz"), like a title in a signature. Later mails in that thread carry no line, because the person already knows.
- Mail a person approved before it went out carries no line, because that person sent it.
- In Slack and Microsoft Teams chats, AUVY posts as an app, and both platforms show it as one.
- When the meeting bot joins, it posts in the meeting chat that it is an AI assistant from AUVY, that it records and transcribes the meeting, and that removing it stops it. If the bot may speak, the first time it speaks it says it is an AI assistant from AUVY.
How AUVY marks what it makes
Images AUVY generates carry a Content Credentials record (C2PA) from the model that made them, which marks them as made by AI. AUVY keeps the original file with that record and shows previews as separate copies.
Each answer in a chat is marked as AI-written in the page's code, and every file version AUVY writes is recorded as written by AI in that file's history. Meeting notes say on screen that AUVY wrote them from the recording and record which model wrote them and when.
Recordings of calls in which AUVY speaks do not yet mark its voice as generated. A mark can also be lost when content is copied, edited, converted or captured in a screenshot, and a mark says nothing about whether the content is correct.
What stays with you
When your company uses AUVY at work, it is the deployer under the AI Act. You tell the people affected by your use where the law requires it. That includes disclosing AI-made images, audio or video that look real, and AI-written text you publish to inform the public on matters of public interest. When you publish what AUVY made, keep its marks. Your company also makes sure that the people who work with AUVY on its behalf know enough about AI for how they use it (Article 4).
What AUVY is not used for
AUVY does not read emotions from voice or faces, does not score people by their behaviour and does not rank job candidates. Meeting notes and the numbers a call shows never rate, rank or judge a person. Our Acceptable Use Policy (§ 6) forbids the practices that Article 5 of the AI Act bans, and allows high-risk uses under Annex III, such as decisions on employment, education or credit, only with a written agreement made beforehand.
Your data and the models
AUVY does not use your content or its outputs to train AI models (product privacy notice, § 5). Which providers run the models, and where, is listed in the subprocessor register.
Questions about this notice go to legal@auvy.ai.