Telling your audience: a notice kit for Vagmi
Effective November 20, 2026
This page is for you, the presenter. It gives you short notices to say or show to the people in your call or room, and a plain explanation of when the law can ask for more.
What each mode does with other people's words
| Mode | What Vagmi uses | Where it goes |
|---|---|---|
| Captions mode (Google Meet or Teams in the browser) | The meeting's live captions, with the speaker names the meeting shows | Most lines that participants say go to the AI provider (Anthropic) through the Vagmi server, to find questions and answer them. No audio leaves your computer. |
| Tab audio mode (a web call without captions) | The sound of the call tab: the voices of everyone in the call | By default, the audio is turned into text on your computer, and no audio leaves it. If you choose AssemblyAI in the settings, the audio goes from your browser to that speech-to-text provider. The text is then handled as in captions mode. |
| Microphone mode (a talk in a room) | The sound that your microphone picks up: you, and people near the microphone | By default, the audio is turned into text on your computer, and no audio leaves it. If you choose AssemblyAI in the settings, the audio goes from your browser to that speech-to-text provider. Lines that look like questions go to the AI provider. |
In all modes:
- Vagmi does not record the call, and its server does not store audio, captions or transcripts.
- The answers show in the side panel of your own Chrome window. If you share that window or your whole screen, people in the call can see the panel.
One-line notices
Say the notice at the start, and post it in the meeting chat or on your first slide, so that people who join late can read it. Use plain words, and do not hide the notice.
Captions mode
"I use an AI assistant, Vagmi, that reads this meeting's captions to find answers in my own material. It does not record the call. Tell me if you would like me to turn it off."
Tab audio mode
"I use an AI assistant, Vagmi, that turns this call's audio into text on my laptop so it can find answers in my material. It does not record the call. Please tell me now if you do not agree, and I will turn it off."
Microphone mode
"I use an AI assistant, Vagmi, that listens through this microphone and turns speech into text on my laptop, to help me answer your questions from my material. It does not record the talk. If you do not want to be heard by it, tell me or the organizer, and I will turn it off."
For a room, also put the notice on the first slide and, if you can, in the event description that people read before they come.
With the AssemblyAI speech-to-text service
If you choose AssemblyAI in the settings, the audio of tab audio mode and microphone mode goes to that service. Use these notices instead.
Tab audio mode:
"I use an AI assistant, Vagmi, that sends this call's audio to a speech-to-text service so it can find answers in my material. It does not record the call. Please tell me now if you do not agree, and I will turn it off."
Microphone mode:
"I use an AI assistant, Vagmi, that listens through this microphone and sends the audio to a speech-to-text service, to help me answer your questions from my material. It does not record the talk. If you do not want to be heard by it, tell me or the organizer, and I will turn it off."
Short chat version (any mode)
"Note: I use an AI assistant (Vagmi) during this call to find answers in my material. It does not record. Details: https://usevagmi.com"
When recording-consent laws can apply
This is a general explanation, not legal advice for your situation.
- Many laws protect conversations, not only recordings. Some laws apply when a third party hears or processes a conversation, even when nothing is stored.
- With the speech-to-text service, a third party hears the audio. When you choose AssemblyAI in the settings, that provider receives the voices of the people in the call or the room. This is the case that is most likely to fall under wiretap and eavesdropping laws.
- One-party and all-party consent. Some places allow a conversation to be heard or recorded when one person in it agrees (you). Other places require every person to agree. In the United States, federal law and most states use one-party consent, but about a dozen states require all parties, for example California, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Pennsylvania and Washington. This list was not checked against each state's law; ask a lawyer. In a call across places, follow the strictest rule that applies to anyone in it.
- Captions mode can also matter. No audio leaves your computer in captions mode, but the words of the conversation go to an AI provider. Some laws cover the content of a conversation, so do not assume that captions mode needs no notice.
- Data protection laws (for example the GDPR in the EU and the UK). Voices, words and names are personal data. These laws usually require that you tell people how their data is used, and that you have a lawful reason to use it.
- Your organization and the meeting host have rules too. Your employer, your client or the event can forbid AI assistants, or require their approval. The meeting platform's own terms and settings can also apply.
- Some situations need extra care. Calls about health, legal matters, money, children, or anything people expect to be confidential. When in doubt, ask first, and do not use the audio modes.
Good practice
- Use captions mode when you can. Turn on an audio mode only when you need it, and turn it off after.
- Give the notice before you turn a mode on, not after.
- If someone objects, stop Vagmi and go on without it.
- Do not use Vagmi in job interviews, exams or assessments.
- Never tell people that Vagmi is not running when it is.
People who have questions about their data can write to support@usevagmi.com. The privacy policy is at https://usevagmi.com.