These terms cover Voke’s AI features. They sit alongside the terms and the privacy policy, which say the same things in more detail. If you have a question, email privacy@voke.co.
1. What the AI features are
- Ask Voke answers questions about your health data in a chat.
- The morning summary is a short note about your recent data, written each morning. Voke’s code picks the readings and works out the trends. The AI only writes a one-sentence summary and two or three suggestions.
Ask Voke and the morning summary are off until you turn them on. Before they turn on, Voke shows you what is sent and to whom, and you agree to it. Voke asks when you set up the app, and you can turn them on later with the “Use AI features” switch in your account settings.
AI apps you connect yourself, such as Claude or ChatGPT, are covered in section 7. They don’t need this switch.
2. What is sent, and to whom
Ask Voke and the morning summary send your health data to Vercel AI Gateway, which passes it to AI model providers. Ask Voke also sends your own last few messages, and nothing else, to TypeSafe AI to check they’re about health.
In more detail:
- For an answer or a summary. Your question, the earlier messages in that chat, and the health data the assistant looks up to answer go through Vercel’s AI Gateway to a company that hosts the AI model. The model is DeepSeek V4.1 Flash, an openly published model. It’s run by hosting companies Vercel works with, not by DeepSeek, the company that made it. Your data isn’t sent to DeepSeek.
- For the topic check. Before Ask Voke answers, the text of your own last 6 messages in that chat, at most, goes through Vercel’s AI Gateway to Jev, a model run by TypeSafe AI. It never gets Ask Voke’s answers, health data from your account or your account details. The morning summary doesn’t use this check.
The health data the assistant can look up is the data you’ve connected to Voke: sleep, heart, activity, workouts, vitals and readiness, and your journal if you keep one. It can also look up what people in your circles share with you, but only where they’ve turned on “Share with their AI” for you.
The list of service providers names every company that handles your data, and where.
3. Where it’s processed
- Every request for an answer or a morning summary asks to be run in the United States. If the answer says it was run somewhere else, Voke throws it away. By then your request has already been sent.
- Vercel, which passes the request on, may process it in the United States and other countries where it operates.
- The topic check can’t be limited to a region in the same way, so we can’t promise where it runs.
4. What is kept, and for how long
At Voke:
- Ask Voke chats are kept for 30 days after each message, then deleted.
- Morning summaries are kept until you turn AI features off or delete your account.
- Voke doesn’t keep the health data the assistant looked up. It reads it again each time.
- Chats and summaries are stored encrypted with your account’s own key.
At the AI companies:
- Every request asks Vercel’s AI Gateway for zero data retention, so it’s sent only to hosts that Vercel lists as keeping no copy. Vercel says the gateway itself keeps no prompts or answers and deletes them once the request is done.
- If no host that keeps no copy can take a request, it fails instead of going somewhere that would keep it.
- Vercel says some hosts’ policies have narrow exceptions, such as keeping data for a short time to check for abuse or where the law requires. So this is what Voke asks for and what Vercel’s agreements with the hosts say, not something Voke can check for itself.
5. No training on your data
Every request also asks that your data isn’t used to train AI models. Vercel says every host it lists as keeping no copy also doesn’t train on prompts. Voke doesn’t use your health data to build or train AI models either.
6. AI isn’t medical advice
Ask Voke answers and morning summaries are written by an AI model. They can be wrong, and they aren’t medical advice. Voke isn’t a medical device. Check anything important with a doctor, and see the health disclaimer.
Answers are plain text. Voke strips links and images from them.
7. AI apps you connect
You can connect your own AI app that supports the Model Context Protocol (MCP), such as Claude or ChatGPT. You approve each one on its own Allow screen on your phone, where you choose which categories it can read. The connection is read-only. You don’t need to turn on Voke’s AI features to connect one.
That app is run by another company, under your account with them. Its own terms, privacy policy and data retention apply, not these. Voke’s zero data retention request doesn’t apply to it. You can disconnect it at any time in your account settings, and it stops working straight away.
8. Turning AI off and deleting chats
To delete one chat, delete it from your list of chats in Ask Voke.
To stop everything, turn off “Use AI features” in your account settings. Turning it off:
- stops Ask Voke and the morning summary;
- disconnects every AI app you connected;
- deletes your Ask Voke chats and morning summaries;
- deletes other people’s Ask Voke messages that used your data; and
- stops everyone else’s AI, including Ask Voke for people in your circles, from reading your data.
You can turn it back on later. Deleted chats don’t come back, and you’ll need to connect your AI apps again.
Database backups are kept for up to 35 days. Chats and summaries in them are encrypted with your account’s key, and deleting your account destroys that key (see the privacy policy).
9. Changes
If we change what the AI features send or who they send it to, we’ll ask you to agree again in the app before they keep working.