Your data.
Where your records come from, where they go, and where they never do — including what each AI does with a conversation after you ask.
This page explains our commitments in plain language. The Privacy Policy is what governs.
Where your record goes. And where it doesn't.
Nobody in between.
Straight from your hospital's own patient-access API into your library. No aggregator, no broker deciding what you see.You're the customer.
You pay us; nobody else pays us for access to you. No ads, no research licensing, no sale of your records.Delete means destroyed.
One click, permanently destroyed within 14 days — and we publish what we keep, for how long, and why.Your family, your consent, your log.
Adults consent for themselves. Every AI read is logged and yours to download. One switch turns AI access off for the whole family.The AI is yours.
Records go only to the assistant you connect, only when you ask — and we never train on them.The full commitments — retention, support access, and what happens if we ever wind down →
Thylib is for when the record is bigger than one person, one hospital, or one chat.
Data layer,
not data broker.
Your records never touch an ad network or a research database. You control what flows where.
Encrypted in transit and at rest
We hold the data so your AI can read it. We don't mine it and we never sell it.Per-account isolation
Every query is authenticated and scoped to you, including inside a family. Support access is logged, and the log is yours.No third-party sharing
Not with advertisers, researchers, or insurers. Ever.Delete anytime
One click; permanently destroyed within 14 days. If we ever wind down, you get notice and an export first.The only place your records ever go is the AI you connect, and only when you ask. What that assistant keeps afterward is governed by its own settings — so we don't just point at the no-training option, we put the exact setting in the answer until you've flipped it. We never see your conversations, and we never train on your records.
Where each AI stands on your data
As of July 2026Where each stands today — and the one setting that keeps your records out of training.
Chats are used to improve models if the model-improvement setting is on — the choice screen pre-selects ON, so check yours. Incognito chats, Claude for Work, and the API are never used.
Settings → Privacy → "Help improve our AI models" — switch this OFF
On personal plans, chats are used for training by default — but you can switch it off. Temporary Chats are never used.
Settings → Data Controls → "Improve the model for everyone" — manually switch this OFF
A sample of conversations is human-reviewed and used to improve the models — for everyone, including paid plans — until you turn it off. Chats a reviewer has already seen are kept up to 3 years.
My Account → Data & Privacy → "Keep Activity" (formerly Gemini Apps Activity) — switch this OFF
You don't have to remember any of this.
Thylib recognizes which AI is asking. The first few times you use one where training is on by default, the answer carries the exact setting above — then it stops, and returns once if that product's policy changes.
Summarized from each provider's published policy, refreshed periodically; their own settings are the source of truth, and only you can change them. Our full privacy policy →
Ready to understand your health?
Start with the demo, then connect your own records free for 90 days — no card, no auto-charge. Delete anytime.