Your data is not our business model.
Every infrastructure and AI company eventually answers a question about your data, even if they don't answer it out loud: is it something to protect, or something to monetize?
For many open source companies, selling your data is the path to monetization. The tools are free because what flows through them isn't. For some platforms that already charge you plenty, the move is subtler: they train on your data to deepen their own moat, and your usage becomes their biggest asset.
We've answered the question differently, and it's a permanent position. LiveKit powers billions of sensitive conversations, from patients describing their symptoms, to customers reading back account numbers, to people saying things to an agent that they'd never type into a form.
All of that data belongs to you and to your users. We will never sell it. We will never train on it without your explicit permission.
Privacy isn't a plan tier or a toggle on our platform. LiveKit will protect your privacy and your data by default.
One of the most critical ways we've committed to this is to ensure that every model in LiveKit Inference is zero data retention, and this post explains exactly what we mean by that and how we enforce it across our own services and every provider we work with.
Expect to hear in the coming weeks about new LiveKit releases that remove the need for training on sensitive, human data to improve everything from expressiveness to turn detection to overall conversation quality. Expect us to talk more about PCI DSS L1 and HIPAA compliance. And we can't wait to tell you about PII redaction. But what you'll never hear about is LiveKit selling or training models on your user data.
We're excited to keep building with you.