Skip to main content
Enclave
Privacy / Self-hosting

How does a private AI assistant become truly private? Self-hosting is the key

A truly private AI assistant comes down to keeping data within your control — open source and auditable, self-hostable, and offline when needed. Enclave lets your private assistant world run entirely on your own machine, with data stored only locally.

Last reviewed:

The short answer

A truly private personal AI assistant = open source and auditable + self-hosted + offline-capable; Enclave does all three.

Key points at a glance

  • "Private" shouldn't rest on terms-of-service promises, but on capabilities you can verify.
  • Three conditions: code is open source and auditable, it can be self-hosted, and it can go fully offline when needed.
  • When self-hosted, Enclave stores data only locally — switch to a local model and it can go completely offline.

"Private" is often misunderstood

Many products claim to be "private," yet your data still lives on their cloud, and whether it's used is something you can only judge from the terms. True privacy isn't a promise but "data never leaving your control" — which requires verifiable capability, not trust.

Three conditions for true privacy

First, the code is open source and auditable: you can personally verify how it handles data. Second, it can truly be self-hosted: data lives on your own machine. Third, it can go fully offline when needed: after switching to a local model and turning off external sync, it sends nothing outward. With all three, privacy moves from "trust" to "verifiable."

How Enclave achieves it

Enclave is MIT open source and auditable line by line, self-hostable on your own machine with a single docker compose; its model layer can be switched to local Ollama / vLLM, and after turning off real-world sync the whole system runs fully offline. Your private assistant world — memory, experts, activity — exists only on hardware you control.

FAQ

  • Is my conversation data safe? Will it be used for training?
    In self-hosted mode, your conversations live only on your own drive — they never leave your machine and are never used by any third party for training. The model layer can connect to cloud APIs like OpenAI and Anthropic, or be swapped for local Ollama / vLLM to run fully offline; you decide which kind of conversation goes to which model.
  • Can Enclave run completely offline?
    Yes. Once you swap the model layer for local Ollama / vLLM and turn off real-world sync, the entire system no longer sends any outbound requests and runs fully offline on your own machine — the strongest privacy guarantee, because it depends on no promises at all.
  • Can I export or delete all my data?
    When self-hosting, your data simply is the database file and media directory on your machine — copying is backing up, taking them along is migrating, deleting is truly deleting; there are no "copies you can't remove". Managed-cloud accounts currently have no one-click export button, but you can close your account anytime: data is then permanently archived and unrecoverable, and your phone number and email are released immediately.

Ready to try it?

Open it in your browser — no credit card, no install.

Related

ShareXTelegramLINEWeibo
How does a private AI assistant become truly private? Self-hosting is the key · Enclave