Your meeting archive is evidence

This is not legal advice. It is a map of where the questions are, written by a software company, and every one of them has a jurisdiction-specific answer you should get from your own counsel. yapyap is a tool. It is not a compliance product and it certifies nothing.

With that said: most people recording meetings have not thought about what happens if somebody asks for the recordings, and the architecture they picked has already decided part of the answer.

A transcript is a document

The starting point that surprises people: a recording and its transcript are records like emails are records. In a dispute, they are discoverable on the same footing as your inbox. "It was just a notetaker" is not a category that exists.

Two consequences follow, and they run in opposite directions.

Verbatim helps you when the dispute is about what somebody promised. A transcript with speakers on every line and audio behind it is a far better answer to "we never agreed that" than anybody's memory. This is genuinely why some people record.

Verbatim hurts you when the dispute is about anything else in the room. Meeting notes are lossy on purpose — somebody decided what mattered. A transcript preserves the joke, the aside, the half-formed idea, the sentence somebody would have phrased differently on the record, and the six minutes of unrelated complaining before the agenda started. All of it is now producible.

An always-on recorder is a decision to keep the second kind of material for years. That may be exactly right for your work — it is much of why journalists record at all. It should be a decision rather than a default.

The service of process you never see

Here is the part specific to cloud tools, and it is the part worth reading twice.

If your recordings sit with a vendor, that vendor can be served directly. A party in litigation can subpoena them, and law enforcement can compel production through whatever process their jurisdiction requires. Whether you are told depends on the vendor's policy, the type of order, and sometimes on a court's non-disclosure requirement that legally prevents them telling you.

So the practical position is: your conversations can be produced from a third party, on a timetable you do not control, in a proceeding you may not be party to. Serious vendors publish transparency reports and commit to notifying customers where they lawfully can, and those commitments are real and worth reading. They are also, necessarily, subject to the cases where they cannot.

The mirror problem is the one from the deletion post: if somebody else recorded the meeting — usually by adding a bot to the call — the copy of your words is in their vendor account, under their employer's control, and neither you nor your counsel has standing over it.

Auto-delete is not a strategy

A lot of organisations respond to all this by turning on aggressive retention limits. Delete everything after ninety days and there is nothing to produce.

That works until the moment it becomes the worst thing in the file. Once litigation is reasonably anticipated — a threat letter, a regulator's enquiry, sometimes earlier than people assume — a duty to preserve attaches, and routine automatic deletion has to be suspended. Continuing to shred on schedule after that point is how a discovery dispute turns into a spoliation argument, and courts have a range of unpleasant responses available.

The practical requirements are unglamorous:

• Somebody must know the archive exists and where it is. Recordings made by individuals on tools they expensed are the classic gap.
• Somebody must be able to stop the deletion, quickly, across every account.
• Somebody must be able to say what the policy was and show it was followed consistently, because inconsistent deletion looks worse than either extreme.

If your recordings are spread across per-seat accounts at a vendor, with individual users free to configure their own retention, none of those three is straightforwardly true.

Privileged conversations

Briefly, and with the hedge repeated: if you are a lawyer, a doctor, a therapist or a priest, adding a third party's software to a privileged conversation raises questions about confidentiality and, in some framings, about waiver. Different jurisdictions treat vendor processing very differently, and a signed processing agreement is doing real work in that analysis.

We are not going to tell you where the line is. We will say that the analysis is much shorter when there is no third party in it, and that this is the single most common reason lawyers give us for wanting the pipeline on their own machine.

What local actually changes, and what it does not

Honesty first, because this is where a vendor is most tempted to oversell.

Local does not make you undiscoverable. A file on your laptop is a document you control and are therefore obliged to preserve and, when properly required, produce. Anyone selling local recording as a way to avoid disclosure is selling you a problem. Do not buy that reading of this post.

What changes is who gets asked, and whether you find out. When the archive is on your machine, there is no third party to serve. A request comes to you, through your counsel, on the record, with the ordinary opportunity to object, narrow, or assert privilege. There is no scenario where somebody else hands over your conversations and cannot tell you.

Legal hold becomes something you can actually execute. The archive is a directory of dated folders, each holding an audio file and a transcript beside it. Stop the deletion, copy it, keep it. There is no vendor retention schedule running underneath on its own clock, and no per-seat configuration to audit across an organisation.

And the duty is now unambiguously yours. No vendor holds it for you, no admin console enforces it, nobody sends a compliance reminder. For a solo practitioner that is a simplification. For a firm it is a policy you have to write, and pretending otherwise would be silly.

The five questions

Take these to whoever advises you, before the archive gets large.

1. Where does the audio physically live, and who else can produce it?
2. If somebody serves the vendor, am I told? What are the exceptions?
3. When a hold attaches, what stops the deletion, and how fast?
4. What is in the export, and is it usable as evidence — timestamps, speaker attribution, original audio, or just text?
5. Am I keeping verbatim recordings of conversations where notes would have done?

That last one is the one nobody asks. The most defensible archive is often the smaller one, and choosing what not to record is a real option that no tool in this category, ours included, will ever suggest to you.