A private AI that doesn’t log you: five things to check

Privacy is not a badge on a landing page. Here is how to verify that an assistant actually keeps your conversations to itself — and what to do when it does not.

Most people assume the worst and stop there. That is understandable, but it is also how you end up using an assistant that quietly keeps everything. A few checks separate the claims from the behaviour.

1. Is history on by default?

If conversations are saved without you asking, that default is what matters. A privacy-first assistant starts with history off, and turns it on only when you choose to.

2. Can you actually delete it?

• One button that deletes a conversation.
• One that deletes everything, without an email to support.
• A clear statement of how long anything is kept when history is off.

3. Is it encrypted — and where?

“Encrypted” only means something with two more words: in transit, and at rest. Ask for both, and for who holds the keys. If the answer is vague, treat it as no.

4. Does it need an account — and how much of one?

The smaller the footprint, the less there is to lose. A sign-up that asks only for what it needs is a good sign; one that wants a phone number and your real name is not.

5. Read the fine print on training

• Your conversations should never be used to train someone else’s model.
• No hidden third-party analytics reading the content.
• Sharing data with partners should be a flat no, not a checkbox.

A private assistant is not one that promises discretion — it is one that has nothing worth leaking in the first place.

None of this is exotic. It is just the baseline, stated plainly, so you can check it instead of trusting a logo.