Recording, Captions and Consent

Two people read this, and they want different things from it.

If you are in a meeting and the recording indicator just appeared: the host has started a recording. What it captures, what Meet Buddy does not capture, where the file goes and what you can ask for are set out below. You do not need an account to be entitled to any of it.

If nothing is recording and you assumed that settled it: it does not. Any participant can switch on captions, which writes down what they say and keeps it against the meeting. That is set out in live captions, and it is the thing this page previously got wrong.

If you host meetings: pressing record is a legal act in most of the world, and in some places it is a criminal one when you get it wrong. Meet Buddy gives you an indicator and an in-room notice. Those are notice. In a good part of the world what the law asks for is consent, which is a different thing, and no software can obtain it for you.

Almost every legal system treats recording a conversation as different from being in one. The systems differ on whose agreement is needed, and that difference is the single most expensive thing to get wrong.

One-party consent

It is enough that one participant — which can be you — knows about and agrees to the recording. This is the rule under the United States federal wiretap statute (18 U.S.C. § 2511(2)(d)), which also removes the protection where the recording is made in order to commit a crime or a tort.

All-party consent

Every participantmust agree. A number of US states work this way — California’s Penal Code § 632, which protects “confidential communications”, is the most frequently cited — and in those places recording without everyone’s agreement can be a criminal offence as well as a civil wrong.

We are not reproducing a list of all-party jurisdictions here, on purpose. Those lists change, the details differ by the kind of communication, and a stale list printed on a page like this one is worse than no list at all, because somebody will rely on it. Check the jurisdictions that actually apply to your meeting, or ask a lawyer who can.

Which jurisdiction even applies

A Meet Buddy room routinely holds people in several countries. Whose law governs a call between a one-party state and an all-party state is a genuinely contested question, and the answer that keeps you out of trouble is not to litigate it: assume the strictest rule that could plausibly apply to anyone on the call.

Outside the United States

The framing is often different rather than looser. In the EU and the UK, a recording of an identifiable person is personal data: you need a lawful basis for making it, you must tell people before you do, and extra rules apply if what is discussed is health, biometrics or another special category. In Australia, listening- and surveillance-devices law is state and territory law and the rules genuinely differ across borders. In Canada, the Criminal Code permits recording where one party consents, while separate privacy legislation governs what a business may then do with the recording. This paragraph is orientation, not advice, and it is not exhaustive.

You can comply with the strictest version of all of the above without knowing which version applies, by doing five things. They cost about fifteen seconds.

• Say it out loud, before you start.“I am going to record this — is everyone happy with that?” An indicator in a corner of the screen is notice; a question is consent.
• Start the recording after you have asked, not before. Then the agreement is captured on the recording itself, which is the only record of it that survives the meeting.
• Mean the question. If someone says no, stop — or let them leave without it counting against them. A request nobody can refuse is not consent in any system that uses the word.
• Say what happens to it. Who will see it, how long it is kept, whether it leaves your company. People agree to a recording they understand and object to one they do not.
• Tell late arrivals too.Meet Buddy puts the notice in the room’s state as well as broadcasting it, so somebody who joins at minute twenty is shown it — but they have not heard you ask, and they have not answered.

Every item here is a behaviour of the software, described so you can check it against what you see in the room.

Only a host can start one

The control is a host control. It is authorised on the server against your live role in the workspace, re-read at the moment you press it rather than at the moment you joined — so somebody whose access was removed during the meeting cannot start a recording with the rights they had an hour ago.

A visible indicator, for everyone

Recording state lives in the room’s shared state, so it is the same for every participant. Nobody sees a room that says it is not recording while it is.

An in-room notice, including for late joiners

The notice is broadcast to the room when recording starts, and it is also written into the room’s state, so somebody who arrives afterwards is shown it rather than walking into a recording nobody mentioned.

A record of whether the notice appeared

Meet Buddy stores the moment the notice actually went out, and stores nothing there if it did not. That field is deliberately not inferred from the recording’s start time: “we started recording” and “we told the room” are different claims, and only one of them is worth having in a dispute.

Recording can be disabled for a meeting

A meeting carries its own switch, so a workspace can run a conversation that cannot be recorded at all rather than relying on everybody remembering not to.

It may not be available at all

Recording needs object storage and a separate recording worker alongside the media server. Where a deployment has not configured them, the control is disabled and says why. It is never a button that appears to work and produces nothing.

This is the part of the product most likely to surprise you, so it has its own section. Meet Buddy has live captions, they are off until somebody switches them on, and switching them on writes down what that person says.

Each person captions themselves

Captions run in your own browser, listening to your microphone. There is no way for a browser to transcribe the audio arriving from other participants, so what gets written down is the speech of whoever turned captions on — nobody else’s. If two people in a room of five enable them, the transcript contains those two.

It is written down and it is kept

Each finished phrase is stored against the meeting as a transcript segment: the words, how confident the recogniser was, when in the meeting it was said, and which participant said it. That confidence reading is used for one thing: stretches the recogniser was unsure of are marked for the people who can already read them, so that a host knows where to check. It is the machine’s opinion of its own hearing, not a judgement of how anybody spoke, and the number itself is shown to nobody. It is not an overlay that disappears when the room closes. It is readable afterwards against the meeting record, by the people who were in the room and by workspace admins — the same people who can reach a recording of it. Everyone else with a seat in the workspace sees that a transcript exists, how many phrases it holds and how much of the room it covers, and not what was said.

A host can take a stretch of it off the record

The record is moderated. This meeting’s host, and admins of the workspace that booked it, can remove a stretch of speech from the transcript — after which the other people who were in the room no longer see it, and a placeholder says a host took it down rather than pretending the stretch was never there. It is not an erasure and it is not private to whoever pressed it. The words are still held, and are still readable by that host and by those admins, so that the decision can be reversed and so that the record cannot be quietly rewritten. Who removed what, and when, is recorded in the meeting’s own activity trail, which everyone with a seat in the workspace can read. Only the retention period below destroys the words themselves.

It is independent of recording, and the room now says so

A meeting that is not being recorded can still be producing a transcript. The recording indicator is about recording and says nothing about captions. There is now a second indicator beside it: while anybody in the room has captions on, the room header names who is being transcribed and how many of the people present that is, and their row in the People panel is marked. It appears only while at least one person has captions running, so its absence means no transcript is being written — and, like the recording indicator, nobody can switch it off for themselves.

Your audio leaves your machine

Captions need a browser feature that today only Chromium-based browsers have, and in those browsers the recognition is performed by a remote service operated by the browser vendor — your microphone audio is sent there to be turned into text. That is the browser’s processing, not Meet Buddy’s, and it is outside this deployment’s control. It is said here because “captions” sounds local and is not.

What it is not

It is not a room-wide caption track, not accessibility captioning of other speakers, and not a complete record of the meeting. A partial transcript that reads as a whole one is its own hazard, which is why the overlay is labelled as your own microphone.

What that means before you turn them on

Treat enabling captions as a decision of the same kind as pressing record, taken for your own voice. Tell the room you are doing it — people who know a meeting is being written down say different things, and that is precisely the interest the law protects. And where the conversation is one nobody would want written down at all, the answer is not to enable them.

This section exists because the absence of a feature is exactly what people assume wrongly, and assuming wrongly here has consequences. Note what is not on this list: speech capture, which has its own section above.

• It does not transcribe a meeting on its own, and it never transcribes other people for you. No server-side speech recognition runs on the room’s audio, and no transcript is produced unless a participant switches captions on for their own microphone — see live captions. Nothing is transcribed for the operator, and no recording is transcribed after the fact.
• It does not caption other participants.Captions are a capability of each person’s own browser, applied to their own microphone. Where a browser cannot do it, Meet Buddy says so rather than showing a control that does nothing.
• It does not analyse who you are. No voice identification, no speaker matching, no sentiment or attention scoring, no facial recognition. Nothing derives a profile from how you sounded or looked.
• A recording cannot run covertly.There is no mode in which a recording runs without the room’s state saying so. Captions are the exception and it is stated rather than buried: a participant transcribing their own microphone is not announced to the room, because it is their own speech being written down and not yours. If that matters to you, ask.
• It cannot stop somebody recording you by other means. A participant can run screen-capture software, point a phone at the screen, or take a photograph, and no meeting product on any platform can detect or prevent that. The indicator tells you about Meet Buddy’s recording, and it is the only recording it can tell you about.

A meeting leaves a record whether or not anybody presses record, and that record outlives the room. It is short, and it is worth knowing.

Attendance

Who joined, when they joined and left, and how long they were present. A participant who drops and rejoins produces several spans, and a total. Hosts can see this per meeting.

In-meeting chat

Messages sent in the room are stored against the meeting, with the sender’s name recorded alongside — so a line still reads even after that person’s account is gone. Treat in-meeting chat as written down, because it is. A private message is stored the same way.It is delivered only to the two people it is addressed to, and it is not shown to hosts — not in the room and not in the meeting record afterwards, which says how many were sent and nothing about what was in them. A host can stop people sending private messages to each other; that switch decides what may be sent and gives nobody anything to read.

Captions, where anybody enabled them

The words that person spoke, kept as a transcript against the meeting. See live captions — this happens with or without a recording, and it is the reason this section exists in the shape it does.

The meeting itself

The code, the room name, when it started and ended, and which booking it belongs to.

Audio and video are not stored unless a recording is running. Attendance and chat are stored either way, and so is any transcript captions produced.

The file is written to the workspace’s own area of object storage, under a path that begins with the workspace’s identifier. That prefix is built by a shared helper rather than assembled by hand anywhere in the code, which is what makes one workspace’s recordings structurally unable to land in another’s.

Meet Buddy stores the object’s key and never a URL. Playing a recording streams it through a route that re-checks, on every request, that the person asking may watch it — so access can be withdrawn by changing who is allowed to ask, and the withdrawal takes effect at once rather than when a link that was already handed out runs down. The one place a short-lived signed link still exists is the programmatic API, where a program asks for one and is told when it expires.

How long it is kept

Nothing here expires on a timer today. Meet Buddy does not delete recordings, transcripts or chat automatically. A recording is kept until somebody deletes it or the workspace is deleted, at which point the workspace’s stored objects are scheduled for removal along with its rows. Where Workflow Corporation, an Iowa corporationor the host’s organisation has promised a retention period — Recordings, transcripts and meeting records are retained until the workspace is deleted or until erasure is requested; no automatic expiry is currently applied. Deleting a workspace removes its bookings, meetings, recordings and transcripts. — that promise is currently kept by a person doing it, not by the software enforcing it, and this page will say otherwise only when that stops being true.

The host’s organisation controls the recording — not Meet Buddy, and not the operator of this deployment. In data-protection terms they decide why it exists and what happens to it, so they are who to ask.

• Ask them to delete it. Say which meeting, and when. Depending on where you live you may have an enforceable right to erasure or to object; even where you do not, most hosts will simply do it.
• Ask what it was for and who has seen it. A host who cannot answer that has a problem independent of you.
• Ask about a transcript too, not only the recording. If anyone had captions on, some of what was said is written down even if nothing was recorded. An erasure request that names only “the recording” can be answered truthfully while a transcript survives, so name both.
• If you cannot reach the host, contact this deployment’s operator at support@workflowcorp.com. They can usually reach the workspace, and they are the right escalation if a workspace is unresponsive. What they can do unilaterally is limited — see the privacy policy for the split.

Objecting during the meeting is also legitimate, and is the point at which it is easiest to fix. Say so, and the host can stop.

By recording a meeting through Meet Buddy — or by enabling captions in one, which writes speech down just as durably — you are confirming that you have whatever agreement the law where your participants are requires, and you accept responsibility for having it. The acceptable use policy makes recording without the consent the law requires a breach in its own right, and the terms put the consequences of getting it wrong on the workspace rather than on the operator.

That is not the operator disclaiming an inconvenience. It is the only arrangement that reflects reality: whoever was in the room, what was said, what was promised about the file and which laws applied are facts that only the host knows.