Security
Security and privacy. Built into every layer.
Your memory is never sold, never shown to advertisers and never used to train a model. This page sets out how it is protected, and the controls you have.
Isolated per user. Each user's memory lives in its own space. A key can be limited to one user, and nothing else can read it.
Encrypted throughout. In transit and at rest, with daily backups kept for 35 days in the same region.
Export and erasure. Export everything held for a user, or erase it: removed at once, and from backups within 35 days.
Scoped access. Every AI app connects through OAuth with its own permission, read-only if required, and can be disconnected at any time.
Your data
What we keep, and for how long.
- What we keep
- Your account, what you add, the memories learned from it (each with the sentence it came from), the questions you ask and their answers, and technical records such as sign-in times and errors.
- Where
- Memory, backups and logs are kept in one region, which the privacy policy names; the models that learn and answer may run outside it. Regulated teams can ask about it in their review.
- How long
- Until you delete it. What you forget, erase or delete is removed at once, and from backups within 35 days. Technical records are kept for up to a year.
- Who reads it
- Nobody at Geniffy, unless you ask us to (to help with a problem) or the law requires it. It is never sold, never used for ads and never used to train AI models.
- The AI that reads it
- To learn and to answer, the text involved goes to Anthropic's Claude models, with Google's Gemini models (Vertex AI) behind them. Neither trains on it. Anthropic keeps what it receives for up to 30 days, in the United States, to check for misuse, unless its usage policy or the law requires longer.
Controls
Every control is self-serve. No ticket required.
See the source. Open any memory to see the exact sentence it came from, and when.
Correct it. Mark a memory wrong and say what is right, and the right one takes its place.
Forget it. Forget one memory, a source, a person or a topic, or erase all of it.
Download it. Everything Geniffy remembers about you, as one file, whenever you like.
Delete the account. With everything in it, from Settings, at once.
Revoke access. Delete a key and it stops working at once. Disconnect an AI app in its own settings whenever you choose.
Engineering
How it is built.
Scoped access. Apps sign in with OAuth 2.1 and PKCE; a key can be limited to one user. Keys and tokens are stored only as hashes.
Isolated memories. Each person, and each of your users, has a separate memory space; a key limited to one space reads nothing else.
On the record. Every API call is kept for 30 days with its request id, so any answer can be traced.
Incident response. A written plan with named roles. If a breach affects your data, you are told without delay.
Who else handles it
Processors.
| Amazon Web Services | Hosting, storage and sign-in |
|---|---|
| Anthropic | The Claude models that learn and answer |
| Google Cloud | The Gemini models behind Claude, and Sign in with Google |
| Resend | The emails we send, such as sign-in codes |
| Razorpay | Payments, for paid plans |
| Google Workspace | Our email, where messages to support arrive |
| HubSpot | Where we answer support messages |
| Sentry | Error reports: what broke and where, linked to an account number only, never to what you add |
| PostHog | The pages you open, linked to an account number only, never to what you add |
Each works for us under a contract and may use the data only to provide its service to us.
Commitments
In place, and on the way.
A status changes only once it is true of Geniffy, never before.
Talk to us
Found a problem? Tell us first.
Report a vulnerability to security@geniffy.com. We confirm we have it and keep you told as we fix it.