When most software calls itself private, it usually means your data is encrypted while it travels and sometimes while it sits on a server. Almost never does it mean your conversation actually stays on your laptop.

Weeve has always meant the more literal version. The part of the app that listens to your meeting, the part that figures out who said what, and the part that writes the summary all run on the laptop in front of you. Nothing about your meeting is sent to us. Nothing about your meeting is sent to anyone.

That has been true since we shipped. What changed in April is something quieter, and we want to be straight about it.

What we removed

For a while, Weeve offered a way out of the on-device model. If you wanted, you could connect your own Anthropic, Google, or OpenAI account inside the app and have one of those services do the summarising work in the cloud instead. It was off by default. Almost no one turned it on. The people who choose Weeve choose it because they don't want a cloud in the loop in the first place.

In April we took the option out of the product entirely. The reason was small and important. As long as that option existed, the answer to "does Weeve send your audio anywhere?" was "not unless you turn it on." Now the answer is just "no."

A practical consequence we hadn't been making enough of: once Weeve has downloaded the models it needs, you don't need an internet connection at all. The summary still gets written on a plane, on a train through the Alps, in a country house with no signal. Several of our users picked us specifically for that. They were right to.

What our servers do see

We want to be specific about what does leave the device. There are four things our servers learn about you:

1. Your account. Your email and a six-digit code at sign-in, so we know it's you.

2. Your subscription. Whether you're on trial, what plan you're on, whether your card is active.

3. A monthly counter. How many recordings you've started, kept as a single number per user. This is how we enforce the public-beta usage cap.

4. Update checks. Anonymous lookups for the latest version of the app.

A separate service we use for product analytics also knows when you start a recording, when you refine a summary, and which model variant you ran it on. It does not know what was in the recording, what the summary said, what your meeting was about, who was in it, or what you named anything.

That's the whole list.

What our servers never see

Your audio. Your transcripts. Your summaries. Your meeting titles. Your speaker labels. Your templates. Your To Do items. Your decisions. None of it. Not encrypted-but-uploaded. Not redacted-then-uploaded. Not uploaded.

Why this is rare

Until recently, building a meeting tool that ran without the cloud meant building one that was either visibly slower or noticeably worse than the cloud version. Newer Macs have changed that. The kind of model that produces a good summary now runs on the kind of laptop most people in offices already own, at speeds that don't make you wait.

So the reason most meeting tools still send your audio to a server isn't a technical one anymore. It's that nobody made them stop.

What this means for you

If you're a therapist, the person sitting across from you doesn't need to wonder whose terms of service their session is governed by. If you advise on deals, the conversation you're protecting isn't sitting in a third party's training pipeline. If you're a lawyer, you don't have a new vendor to add to your privilege analysis.

The simplest version: your meeting content never leaves your device. That's not a marketing line. It's the result of choosing, every time it came up, the version that makes it true.