How Quorum handles your data

Three things are worth stating plainly, because they are what people actually ask.

We do not train AI models on your content

Not ours, and not anyone else's. Customer content and the material Quorum generates from it are never used to train, fine-tune or evaluate any general-purpose model, and are never contributed to a shared or public dataset. Our model and speech providers are engaged on terms that forbid them from training on the data we send.

This is a commitment in the Terms of Service, not a setting.

We do not clone anyone's voice

Briefings are narrated by a fixed pool of pre-made synthetic voices that we license. The platform has no voice capture and no cloning capability, so this is not a control that could be misconfigured.

We only read what we are pointed at

• Calendar access is limited to the Calendars.ReadBasic.All permission. We read basic event details. We do not read mailboxes or message bodies.
• Document search runs over the storage locations you connect, and you approve which files are used before a briefing is generated.

Who is responsible for what

For meeting content, the organisation whose workspace it is decides what happens: it is the controller and Quorum is its processor. If you are an attendee or someone named in a document and you want your information removed, contact that organisation. We will help them act, but we cannot act without them.

Quorum is the controller for a narrower set of things it decides itself: account administration, billing, security, support correspondence and its own website.

Questions about any of this go to privacy@quorumtech.ch.