Tessera
Guide
How to get personal AI answers without giving up privacy
The trade-off is not as fixed as it looks.
Being known makes answers better. Being tracked makes you the product. Those are not the same thing, and the difference is structural rather than a matter of trust.
Why the trade-off feels inevitable
For twenty years, personalisation was funded by advertising. Better answers meant more data, more data meant better targeting, and the arrangement was never really explained. So "it knows me" and "it is selling me" arrived together and now sound like the same sentence.
They are separable. What links them is a business model, not a technical requirement.
The five conditions
1. The company should not need your data to make money
If revenue comes from subscriptions, there is no advertising machine downstream that gets better the more it knows about you. Check how the company earns before you check its privacy page.
2. You should be able to read everything it holds
A memory you can open, line by line, is a memory you can govern. One you cannot see requires trust, permanently.
3. Only what is relevant should travel
Sending your entire profile with every question is both worse for privacy and worse for answers. Relevance is a privacy feature.
4. Your words should not become training material
Being answered and being used to build a product are different things. This should be a contractual term, not a toggle you have to find.
5. Leaving should be possible and complete
Getting a full copy of what is held about you, deleting it for real, and closing the account should all be possible, and they should cover everything rather than a summary. If any of those is missing or partial, the memory is not really yours.
The short version. Personalisation is safe to the exact extent that you can see it, correct it, limit it, and end it. Everything else is a promise.
Sensitive categories deserve more than a policy
Health and money are the two categories people most regret sharing and most benefit from having understood. Those specifically should be encrypted at the level of the individual fact, with keys the operator cannot use to read them, so the protection is a property of the system rather than of the staff handbook.
Where Tessera lands
Subscriptions only, no ads business. Every fact visible and editable. Only relevant facts travel with a question. No training on your conversations by default on the business tiers we use. A full copy of your data on request, and real deletion. All of it explained, and the binding policy if you want the legal text.
Questions
Related questions
Can AI be personal and private at the same time?
Yes, if five things hold: the company does not make money from advertising, you can read everything it stores, only relevant details travel with each question, your words are not used for training, and you can export and delete everything.
Is it safe to tell AI personal things?
It depends on who holds the record and what they may do with it. The safest arrangement is one where you can see every stored fact, delete it for real, and where the business is funded by subscriptions rather than advertising.
Should health information go into an AI app?
Only where sensitive categories get more than a policy promise. Look for field-level encryption with per-user keys, so those facts are unreadable even to the people operating the service.
Stop explaining yourself to AI.
Ten days free, no card required. iPhone and web.