Short enough to read twice, and worth reading twice. Then we change a couple of settings together and you are done.
Do this
- Read the list. Never type in: passwords, bank or card numbers, your Social Security number, your full date of birth alongside your full name, account numbers, or anything from somebody else’s private records.
- Notice what is not on that list. Your symptoms in general terms, your questions, your drafts, your worries, the contents of an ordinary letter. Those are all fine, and being frightened of the whole tool costs you more than it protects you.
- Now open your assistant’s settings. Look for Data controls, Privacy, or something similar.
- Find the setting about using your conversations to improve the model, and turn it off if you would rather they did not.
- While you are there, look at whether it saves a memory about you between conversations. Read what it has stored. Delete anything you would rather it did not keep.
Watch it done first, if you would rather
Let my human assistant John show you what that looks like. He does the same thing you are about to do, at the same speed, with nothing edited out.
Filming This one is not recorded yet. Everything you need is written out above, and the video will appear here when it is ready.
What just happened
The short version of what happens to what you type: it goes to a company’s computers, it is stored, and depending on a setting you just looked at it may be used to train future versions. Staff can in principle see it, usually only for safety or abuse investigations.
None of that is sinister, and all of it is ordinary for an internet service. It is the same shape as email. The reason for the list in step 1 is not that these companies are villains. It is that anything stored anywhere can eventually leak, and a password or an account number is the kind of thing that causes real harm when it does.
The memory feature is the one that surprises people. Some assistants quietly build up a profile of you across conversations, which is genuinely useful and worth knowing about rather than discovering later.
If you did the blanking out at stop 5, you have already been doing this properly. This stop only names what you were doing.
If something did not go as expected
- I cannot find the settings.
- Look for your initial or a small circle in a corner, then Settings. It is usually two clicks from there. Or ask the assistant itself where the setting is.
- I already typed something from that list, weeks ago.
- Delete that conversation, and if it was a password, change the password. Then stop worrying about it. That is the whole remedy.
- Is it safer to pay for it?
- Paid versions sometimes default to not training on your conversations, which is a real difference. It changes nothing about the list in step 1.
The view from here: What these companies keep